#trilema goes dark

Friday, 02 August, Year 11 d.Tr. | Author: Mircea Popescu

For the record, here's its last day (as seen by me) :

**** BEGIN LOGGING AT Fri Aug 2 08:09:49 2019

Aug 02 08:09:49 * Now talking on #trilema
Aug 02 08:09:49 * Topic for #trilema is: If you don't know where you are, you shouldn't be here.
Aug 02 08:09:49 * Topic for #trilema set by mircea_popescu!~mircea@pdpc/supporter/silver/mircea-popescu at Wed Sep 5 23:57:45 2018
Aug 02 08:09:50 -ChanServ- [#trilema] To speak in #trilema you must be voiced. If you have a RSA key registered with deedbot, send !!up to it in a private message, decrypt the challenge string and return it with !!v ; else politely ask one of the voiced people to voice you.
Aug 02 08:09:50 * #trilema :http://trilema.com
Aug 02 08:10:18 * deedbot gives voice to mp_en_viaje
Aug 02 08:12:08 mp_en_viaje http://btcbase.org/log/2019-08-01#1926141 << understand, the relationship is not symmetrical. this "mod6 isn't building off what mp says, just says his own things" thing you're trying will not work out any better for you than it worked every other time someone else tried it before. i get that the self-feeling of specialness is the core and root of all hominids, but that doesn't promise it also does something outside the skull. se
Aug 02 08:12:08 mp_en_viaje lf-soothing as it may be.
Aug 02 08:12:30 mp_en_viaje you're currently in this position where you've neither done anything useful nor said anything intelligent in years ; and i will send you packing the next time this comes around.
Aug 02 08:12:44 mp_en_viaje derping about "as if you believed otherwise" is not a workable strategu.
Aug 02 08:13:51 mp_en_viaje if you don't know what you should be doing, you don't belong here. if you do know what you should be doing, start fucking doing it already.
Aug 02 08:16:14 mp_en_viaje o look at that, phf logger also quit sometime last night. hey!
Aug 02 08:17:18 mp_en_viaje http://logs.bvulpes.com/trilema?d=2019-8-2#475987 << actually atm i'm stuck reading bv's logotron, because guess what!
Aug 02 08:17:56 mp_en_viaje oh, and hm, wtf, this has no readouts.
Aug 02 08:18:37 mp_en_viaje http://log.mkj.lt/trilema/20190626/#101 << scriba ended logging on the 26th, so that's not useful either
Aug 02 08:18:38 scriba Logged on 2019-06-26: [17:45:27] iirc there exists a 'fetish' set who dream of... being eaten by cannibals. i never grasped exactly how this worx, psychologically, but entirely possib that 'is proof of my deliciousness!'
Aug 02 08:20:32 mp_en_viaje http://blog.lobbesblog.com/2019/07/time-estimation-for-logotron-based-off-of-logbot-august-18th/ << i guess we're stuck with this.
Aug 02 08:21:19 mp_en_viaje btw lobbes for some reason your root (http://lobbesblog.com/) doesn't load.
Aug 02 08:24:31 mp_en_viaje http://logs.bvulpes.com/trilema?d=2019-8-2#475954 << this has been ongoing for a while
Aug 02 08:24:37 mp_en_viaje also, this newmode of idiocy does not in fact work.
Aug 02 08:25:37 mp_en_viaje i will be setting the #trilema channel to silence until we actually have working logs ; which is hopefully mid august. this situation where there's a single logger that's even still going coincidentally, perhaps because maintainer was so deeply unreachable others couldn't even reach out to sync downtime is too much to paper over.
Aug 02 08:26:06 mp_en_viaje in the meantime, everyone's invited on trilema & other blogs.
Aug 02 08:26:32 * You are now known as mircea_popescu
Aug 02 08:26:42 >chanserv< OP #trilema
Aug 02 08:26:43 * ChanServ gives channel operator status to mircea_popescu
Aug 02 08:26:58 feedbot http://trilema.com/2019/holy-shit-music-videos/ << Trilema -- Holy shit, music videos
Aug 02 08:28:52 * feedbot has quit (Write error: Connection reset by peer)
Aug 02 08:34:50 * mircea_popescu removes channel operator status from deedbot
Aug 02 08:34:54 * mircea_popescu removes voice from whaack
Aug 02 08:34:58 * mircea_popescu removes voice from trinque
Aug 02 08:35:01 * mircea_popescu removes voice from spyked
Aug 02 08:35:08 * mircea_popescu removes channel operator status from scriba
Aug 02 08:35:13 * mircea_popescu removes voice from scriba
Aug 02 08:35:17 * mircea_popescu removes voice from phf
Aug 02 08:35:20 * mircea_popescu removes voice from mod6
Aug 02 08:35:23 * mircea_popescu removes voice from Mocky
Aug 02 08:35:26 * mircea_popescu removes voice from lobbes
Aug 02 08:35:30 * mircea_popescu removes voice from jurov
Aug 02 08:35:33 * mircea_popescu removes voice from dorion
Aug 02 08:35:37 * mircea_popescu removes voice from diana_coman
Aug 02 08:35:41 * mircea_popescu removes voice from danielpbarron
Aug 02 08:35:44 * mircea_popescu removes voice from bvt
Aug 02 08:35:48 * mircea_popescu removes voice from BingoBoingo
Aug 02 08:35:51 * mircea_popescu removes voice from billymg
Aug 02 08:35:55 * mircea_popescu removes voice from ave1
Aug 02 08:35:59 * mircea_popescu removes voice from auctionbot
Aug 02 08:36:03 * mircea_popescu removes voice from asciilifeform
Aug 02 08:36:09 * mircea_popescu removes voice from mircea_popescu
Aug 02 08:36:14 * mircea_popescu removes channel operator status from mircea_popescu

And well...

Category: Bitcoin
Comments feed : RSS 2.0. Leave your own comment below, or send a trackback.

176 Responses

  1. As asciilifeform points out, there are some dubious patterns with logotrons - if not even wider, with tmsr infrastructure development and maintenance as such and I don't really understand why original maintainers seem to just silently drop stuff and vanish/get eaten or what??

  2. For that matter now I can't even subscribe with deedbot to the comments feed for this article, deedbot says "failed".

  3. I picked up this grenade today. Would like to put some of the incremental barf here, in case I get tripped somewhere along the way:

    (1) Phf's raw .txt dumps (scraped off his ghost ship this morning) :
    http://www.loper-os.org/pub/logism/phf.tar.gz
    (SHA512 :
    a69e860af8453b07e19dbeec47161a35a0593497c724f9d45ec17fd6d2d866d5e79772af0994871e9d83182be2e4576f120fba80260ad969434c21e322a781a9) and his orig copy is still (as of the time of writing) up. For now.

    (2) DB schema ("postgres" flavoured, to utilize phuctor-style O(n log n) text search) :
    http://www.loper-os.org/pub/logism/schema.sql.txt

    (3) Eater which correctly pumps (1) into (2).
    http://www.loper-os.org/pub/logism/eat_dump_py.txt

    (4) Viewer is ~80% written, however it does not yet replicate Phf's backlink functionality.

    (5) IRC backend is to be the previously published CL item.

    Will genesis 2-4 when complete, and stand up on Dulap.

    If all of this helps someone get to the finish line before I do, this would not be a bad thing -- by all rights there ought to be multiple (documented!!) loggers.

  4. Well this explains why it was so quiet today. About to get head home from the mines and then I'ma continue cranking out that php logotron. If this ain't motivation I dunno what is.

    Btw root of lobbesblog.com is dead due to it being hosted on that heathen Frantech vps. I'm about to tell them to fuck off (they gladly took my $5 for this month but... no moar service.)

  5. Also http://nosuchlabs.com/pub/snapshot_signed.txt ( and would like people to replicate and compare, while the orig. remains up...) .

  6. Mircea Popescu`s avatar
    6
    Mircea Popescu 
    Friday, 2 August 2019

    @Diana Coman Trilema is both designed and intended to work as an elevator, taking anyone from the ground floor up to all the floors. People however are very systematic about "only moving among this many floors". Whole elevation history was therefore this tedious exercise in "whenever floors change prefix, a set of people will get off". Because if he went 2014-2016 gotta stop there, can't possibly keep going 2016-2018, 2018-2020 etc.

    At least that's been broadly the experience to date -- education is not so much about actual education, but rather modulated expression, as in genetics. For as long as what's outside matches what was inside anyway dude's in for the ride, and once the match falters that's it, dude's wandering off into the background. In any case out of the fucking question the insides get revised or anything scary like that ; much better to just wait out reality. Who knows, maybe it alligns with the insides again ? Surely it must be cheaper for the whole universe to alter itself, rather than demand frog grows fur or some other inconceivable in the direction of actual development. Nature moves by natural selection, after all, nobody teaches the bacteria to become plants, there's no school of evolution. Why should people be any different ? Not like the brain actually is what it pretends itself to be.

    Also, afaik it was feedbot that did feeds/subscription, not deedbot ?

    @Stanislav Datskovskiy Thanks for the log, those are good to have ; I'm gonna put a copy up myself in a minute. Since the chan is silenced, we even have guarantee of completeness. Did you spot check for correctness at all ? Might be a good idea.

    I'd much rather have multiple loggers also.

    @lobbes Cheers.

    PS. I took so long to say anything because also travelling today, not because being dramatic. So hi from Warsaw, and teh Republic prevails.

  7. > Did you spot check for correctness at all ?

    Indeed did, and superficially seems to be accurate copy, running up to when the bot fell down (the 'raw' .txt dumps however all post-date "dragon", the earlier material will have to be extracted from the htm in the 2nd (larger) tarball.)

    And it won't hurt to verify that the result lines up with what people have in their local log stashes. This is somewhat easier to do after parse into DB.

  8. I will definitely be cribbing from Stan's notes here (ty @Stanislav Datskovskiy for grabbing this grenade). I'm currently running a scrape of btcbase using the method outlined in http://nosuchlabs.com/pub/snapshot_signed.txt. I will compare and sign once complete.

    Agreed there is no harm in multiple documented loggers. Once I get my thing published I will also spin up loggers for #trilema and the other castles on request.

  9. I can totally see the elevator. And sadly I have also repeatedly seen (irl too) the "can't possibly change" - I don't have any shred of "what do" or how to go about it. I suppose it's also true that anything has at best a limited ability to change - so ultimately at some point everyone and anyone does fall out of the life-elevator if of nothing else. The only bit I am rather hazy about is whether/why there doesn't seem to be much of intermediate/slower levels as it were (perhaps I can count #ossasepia as providing some of this too) - irl at least to some extent there is some possible amount of more graceful "getting off" so honest people that admit it's too high for them may take a step back and still be a smaller but still useful part of it.

    Re rss, right you are, feedbot works. Although deedbot's help still lists !!subscribe so I don't know if it's still supposed to work but borked or it's just decay starting to show there too.

    Stan, thank you for the logs, I grabbed the file and I'll have a look at it this weekend. Once there is *some* logger available, I plan to run it too as I think it should be pretty much a basic bar even more basic than running a node.

  10. Mircea Popescu`s avatar
    10
    Mircea Popescu 
    Saturday, 3 August 2019

    @Stanislav Datskovskiy
    > the earlier material will have to be extracted from the htm in the 2nd (larger) tarball.)

    The source of all that material is a tar I gave him at some point cca 2017 (it was a discussion in the logs, I could even link to it, in principle).

    I can definitely reproduce it, though not conveniently while travelling. So that could be an alternative pathway, though not active until ~end of year.

    The other sticky point is that I actually own the btcbase domain. I say sticky because, although I could redirect it to a logger (which then, if it were to use the same url format, would require no editing of the thousands upon thousands of url references to log lines), I'd have to redirect it to a single logger, which seems to me undesirable.

    Could I set the domain to round-robin the whole list of loggers I wonder ? Such that btcbase.org/log/ would then point to a /log/ page maintaiend by each log maintainer, a page which works compatibly, but which may include a topline like lobbes' does, "this is X's log, you can also p, q, r, s pages/projects" ?

    @lobbes Sweet.

  11. Mircea Popescu`s avatar
    11
    Mircea Popescu 
    Saturday, 3 August 2019

    @Diana Coman Lol simulcomment.

    > I don't have any shred of "what do"

    The only "what do" I can think of is discussing the matter explicitly.

    Let man be monkey if he wants to, but let him then be monkey explicitly, in the face of clear and neat textual description of his being monkey.

    Let him at least have to put in the effort to ignore, and to pretend, against clearly spelled out reality, rather than indistinct implication. Let this little wall stand, as easily jumped over as it is, less than a foot tall. Let man have to confront the shame, and the guilt, of "no, you didn't lose the way ; you explicitly jumped over", for all the good it'll do ; and even if it won't do any good at all.

    I have no doubt that sufficiently dedicated man (and ~all monkey) will have little trouble making it over the one-brick-tall wall, especially if they set their "mind" to it. But I'm not doing it for them. I'm doing it for me.

    > it's also true that anything has at best a limited ability to change

    I do not know that this is true. Everything has an unlimited ability to change ; though rate of change (ie, change/time) is universally limited. But then again, there's a difference between "child will not grow any faster than child grows" and "anal child will not grow ''because anyway''". In fact, it's not so much that there's a difference, it's rather that there's no similarity.

    > at some point everyone and anyone does fall out of the life-elevator if of nothing else

    So why die any sooner than one has to.

    > so honest people that admit it's too high for them may take a step back

    The problem is psychological, not ontological, which is why and wherefore we have a buncha lords we keep having to shave (and the shavings dissolve into nothingness, suggestively "as if there were nothing there to begin with"), rather than a teeming sea of knights we keep finding ourselves forced to promote.

    One could say this is due to the "particulars", ie "mp's fault", but then the whole process changed towards castle promotion a while ago and... little changed. The substance of the problem's not objective limitation, but learned, and deliberate, self-castration.

    > even more basic than running a node

    In normal time. In crazy time it's not more basic because failure with former is private whereas a shitty logger's shittiness is publicly visible.

  12. > If all of this helps someone get to the finish line before I do, this would not be a bad thing -- by all rights there ought to be multiple (documented!!) loggers.

    As soon as I finish documenting that Hunchentoot ball of yarn (early September), I'll also write, genesize, publish etc. a logger code as a demo. I expect it'll lack many of the btcbase features, my idea would be to write this up as a basic didactic example, so that whoever's interested writing a logger would understand what they require.

    > Could I set the domain to round-robin the whole list of loggers I wonder ? Such that btcbase.org/log/ would then point to a /log/ page maintaiend by each log maintainer, a page which works compatibly, but which may include a topline like lobbes' does, "this is X's log, you can also p, q, r, s pages/projects" ?

    Btw, is there a #trilema logger spec somewhere? I guess Phf's implementation could serve as a reference, but there are parts such as the search functionality whose workings aren't immediately obvious.

  13. Mircea Popescu`s avatar
    13
    Mircea Popescu 
    Saturday, 3 August 2019

    By tradition (and because I insistently believe this is the right way -- such that we end up with short and simple standards sanctified by lived experience, whom everyone is willing to defend, rather than lengthy and prolix standards everyone's willing to shave parts off) we spec #trilema things as the need arises, rather than at the onset.

    Consequently there isn't much in terms of logger specs. What little there is I'll reproduce below (seeing how it'll take me five minutes to do and anyone else'd be stuck digging through logs for months).

    1. Logger (like all bots) MAY NOT speak unless spoken to ;
    2. Logger (unlike other bots) MUST read out the content of the log it maintains, if the previous speaker isn't itself (even if it is another logger bot), and if the previous speaker's line included a logline reference to that log.

    Other than that bot behaviour, there's some soft standardization of the log url format, in the sense that we seem to not mind the log/YYYY-MM-DD#LINECOUNT format for single-line reference. Just prior to Framedragger's disappearance we were discussing multi-line selection, which he also implemented (as log.domain/CHANNEL/YYYYMMDD/from:START/to:END#ANY). I happen to like it, but it never became anything owing to his disappearance (and yes I was intending to push everyone towards the format -- minus the datedump, I think YYYY-MM-DD is actually superior to YYYYMMDD). I do not see the value in shaving two bytes by using YY-MM-DD, as this thing is very rarely typed by hand, and even if so typed, usually as an insert not de novo.

    If we were to decide on log.domain/CHANNEL/YYYY-MM-DD/PROGRAMMABLE I expect Trilema references to loglines could be rather easily regexp'd into correctness, so it wouldn't be the humongous sort of cost that say selection standardization became.

    Other than this, I believe "full scope windows", such that lines run the width of the user's display whatever it may be, actually is the whole of the law ; though alternating background for lines, of faint colors of which one is the whole page background and another slightly lighter (as displayed on mkj, not on btcbase or bvulpes) is a great win for readability ; I also believe it helps selection backing stand out if the page background isn't straight white but something more like #EFEFEF / #FDFDFD as seen there. (#ffa happens to make excellent underscoring, btw!)

    As far as search is concerned, the desired behaviour is (I think), as follows :

    1. Strings are to be matched exactly, including capitalization and not skipping any non-ascii characters, (so that text matches none of tExt, téext, e-text, texte or texté)
    2. Quoted strings are to be matched approximately, excluding capitalization and skipping any non-ascii characters, (so that "text" does indeed match any of tExt, téext, e-text, texte or texté)
    3. f: will match lines issued by the respective name (same rules for the string match as above), and we will need a mechanism to group multiple names into a single search name (this was discussed on the 19th, "i wonder, did i say mircea_popescu ? mp_en_viaje ?)
    4. This was never discussed before, but I suspect we might want ()|& added, such that (a|b)&c matches if c and one of a or b match. There's a definite usecase for it in my own mind, and it might be advantageous from the pov of running the thing to have fewer searches run by using this more complex format than more run on a simple format. This would also resove the straggler issue with point 3 above, I could just (f:"mirc"|f:"mp_")&(fuck&("cunt"|"snatch")) for great ejaculations. I'm sure you're just as glad to read about yet-another-ad-hoc-scripting-engine as everyone else. Wut can we do ?!

    That about covers it, I expect. Comments very much welcome!

  14. Mircea Popescu`s avatar
    14
    Mircea Popescu 
    Saturday, 3 August 2019

    PS Took me like three quarters of an hour rather than 5 minutes, but what can you do.

  15. > Strings are to be matched exactly

    Currently btcbase does some preprocessing on strings before matching them, which I personally find useful. For example http://btcbase.org/log-search?q=steal+from matches both "steal from the dumb to give to the smart" and "steal shit from the house", but also "from the period, adults pretending to steal kids' noses". As far as I can tell, it:

    a. splits the search input into tokens, by a separator character (or are there more?); and
    b. returns the strings which contains all the tokens, regardless of the order of appearance.

    Meanwhile:

    > Quoted strings are to be matched approximately

    This also seems to have another useful meaning in the current implementation. If we quote the example above, then http://btcbase.org/log-search?q=%22steal+from%22 will match "steal from the dumb", but it won't match the other strings anymore, which means that quoting prevents that tokenization (splitting into "search terms"?) from taking place.

    > I suspect we might want ()|& added

    In light of discussions about finding things (blog archives, stuff on the internet etc.), these might turn out to form more than just an ad-hoc scripting language, if said language later materializes into a general "find on stuff on the 'net" search mechanism. I can definitely see the merit, I'm curious to hear other opinions. Speaking of which, should this thing contain a negation operator?

    OTOH, the only nuisance I see is that this introduces special syntax into the search, so now if I want to search for the string a & b, I'll have to quote the & (or the whole string) so that it doesn't get interpreted as a search operator.

  16. Mircea Popescu`s avatar
    16
    Mircea Popescu 
    Saturday, 3 August 2019

    The functionality you discuss re ?q=steal+from is logically equivalent to steal&from, neh ? There's nothing wrong with implicit-&, if you want to pre-process the nude natural language sentence hurr durr and qurr into hurr&durr&and&qurr. Maybe I should edit (or rather, restate) the spec above to include this, I guess.

    > now if I want to search for the string a & b, I'll have to quote the &

    There absolutely will be no escapisms ; as the price for that I can live with "the only way to search for strings that include & | ( or ) is to use the quoted string format ; plain string won't yield anything". So you can do "a & b" which will match "foca & bmo". If you tokenize quotes first there won't be any problem with "a & b"&B, either.

    And yes, there's very much a view towards "find human text in the general" contemplated here.

    I do not believe a negation is either useful or desirable, in light of the very large sets we're dealing with.

  17. Mircea Popescu`s avatar
    17
    Mircea Popescu 
    Saturday, 3 August 2019

    PS. I like today's header so much... I updated my avatar!

    This is the first time, since 2008!

  18. @Mircea Popescu :

    The "round robin" thing sounds appealing, but I suspect that for some time the various loggers will have mildly differing feature sets, and this will cause confusion. Plus ideally we'd like less reliance on DNSism, rather than more.

    Further re: "btcbase" -- at some point will have to replicate Phf's vtree viewer -- possibly a somewhat harder nut to crack than the log.

    Re: search functions: currently in my draft I have a working search, but it is Phuctor-style, i.e. relies on more or less unadorned function provided by Postgres.

    I will be genesising the whole orchestra, however, so the item can be brought to spec with help of other folks.

    In particular, will need to ask Trinque to cough up his anti-fleanode-wedge patch for the IRC backend; and will need to add echo functionality (as outlined in #13), IRCtronic search/"seen"/etc commands, to same; currently it is "mute".

  19. Mircea Popescu`s avatar
    19
    Mircea Popescu 
    Saturday, 3 August 2019

    The trigrams thing ? It did seem pretty appealing last it was discussed.

    In truth, getting good quality search going is not trivial ; probably the heaviest part of this whole thing. BV's logotron (which, in fact, is the only one still functional) does actually offer search (though I dunno he ever published ; and it is of a different flavour than both btcbase and what's contemplated here).

    The dns thing... I dunno. As long as we're still using domain names at all, using them intelligently rather than unintelligently isn't necessarily a crime ; having different logotrons is in a sense good, but then again standardization is also good... nfi.

  20. Fwiw I ran the wget (with additional 1 sec delay between requests + output to file) and there were no errors I can see. Comparing to Stan's archive, the log-raw dir is identical (according to diff) while in logs the differences are in html comments, specifically "" vs render 0.339s.

    Manual spotcheck between btcbase.org/log and what I got from wget was fine too.

  21. Ugh, still ate the html bit: render 0.336s vs render 0.339s in html comments in the html (i.e. not the raw) log files.

  22. Mircea Popescu`s avatar
    22
    Mircea Popescu 
    Saturday, 3 August 2019

    The idea is to check against your own local logs. Mine are atm in deep freeze, which is why I'm oversensitive I guess.

  23. I did a few spot checks against my local logs too and I couldn't find any mismatch (obviously, other than the fact that local log goes beyond the point where btcbase.org stops) but I don't know that this in itself says much really.

    Perhaps part of trouble with "different and without standard" is this that it can make it more difficult to compare the content directly.

  24. Salutations lords & ladies & gents. I had planned to introduce myself after at least making it through the July log, but it looks like there's no time for "no time like the present" like the present! My key is at http://welshcomputing.com/jacob-welsh.asc (25999E5095BA9A2E1CE6305EA8902A7CACFD13B0, 198.199.79.106). S.NSA may also have it from a FUCKGOATS order (happy customer btw, A++).

    @Stanislav Datskovskiy - I have mirrored your snapshot: sig, payload.

    I also did a wget of phf's blog: http://welshcomputing.com/files/barksinthewind.com.tar.gz. I did not use --convert-links, but did manually fix up some links to 107.170.141.103, which was included in the domains, and scrubbed some speshul-custom-just-for-me hidden inputs.

    On the interchangeable loggers question, is there a solution to the message number divergence problem? That is, even assuming a well-behaved server, eventually one bot is going to have an outage and miss some messages so A's #123456 is B's #123460 and so on. Would also happen when a bot isn't voiced so is sending readouts but not being heard, as appears to have happened to mimisbrunnr. Even in gossipland as I understand it, nodes will learn of messages in differing order. Only solution I know of to this problem is Bitcoin, which, well...

    Does the fact that & is the delimiter for URL GET parameters argue in favor of + as the conjunction operator for search?

    I have some homemade Python logbot/www parts on the shelf from a past project. I plan to give them a re-read, determine what would be needed to bring them up to the standards of the Republic, and report back. Rough guess is two days to a week after that to get something online meeting the basic requirements except for search and pre-dragon history. However I'm unsure if my running such a thing would be desired at this time given my newness and @Mircea Popescu's "shitty logger's shittiness is publicly visible" point. Fwiw, I would publish the code and ~complete~ raw data, and I don't plan on evaporating any time soon and would give notice if at all possible should this change.

    Otherwise, getting a MPWP up is a high priority for me; I have a number of things to begin sharing that I think will be of interest.

    @lobbes - in my view Frantech is not merely heathen but an outright scam. Perhaps we were lured by the same bait of "takes Bitcoin!" Oh yes, they took the coin just fine. I can share the full lulz of the educational experience if anyone cares.

  25. Sneak preview : http://logs.nosuchlabs.com/log/trilema/2019-08-01

    It doesn't log yet! (still gotta plug in backend) and will flicker in and out as I saw on it. And a number of important knobs ("from:", search pagination) missing. But does display Phf's logs, near as I can see, correctly.

    Not genesis'd yet, but full source can be seen here: http://p.bvulpes.com/pastes/PHssz/?raw=true

  26. Mircea Popescu`s avatar
    26
    Mircea Popescu 
    Sunday, 4 August 2019

    @Stanislav Datskovskiy I'm getting 502 bad gateway ftr.

  27. Err, http://p.bvulpes.com/pastes/ZBanJ/?raw=true (prev. paste was broken ver), if anyone cares.

  28. @Mircea Popescu should be up just now. I'ma off to eat, will read commentary barf (here and in #a) when come back.

  29. Also will post the templates (realized that they aint in the paste).

  30. Mircea Popescu`s avatar
    30
    Mircea Popescu 
    Sunday, 4 August 2019

    @Diana Coman As long as nobody notices disparities I'm happy.

    @Jacob Welsh

    > Does the fact that & is the delimiter for URL GET parameters argue in favor of + as the conjunction operator for search?

    Yes, absolutely. As it turns out Ima definitely have to restate that spec draft at #13.

    > I'm unsure if my running such a thing would be desired

    Sure, as long as you stick around to learn from it.

    > is there a solution to the message number divergence problem?

    Yes -- if your bot fucks up you put in the elbow grease and fix it, by hand. Plenty of incentive to not make bots that fuck up right there.

    Gossipd is eminently not an applicable analogy here, irc is specifically built around central server. A "if I go to town I expect fine dining and well polished sluts ; if I go to countryside I expect excellent tomatoes and quiet wooden bench under nut tree -- do not try to sell me on the wooden bench in the ~relatively~ quiet spot in town or the ~relatively~ fine dining of the countryside" sorta situation, if I accept the central server you'd better never mention the clock as a "problem".

    Certainly having your blog is a great boon -- consider how helpful it turned out to be for me right here!

    @Stanislav Datskovskiy Looking good!

  31. @Mircea Popescu , @Diana Coman : still trying to think of a reasonable (i.e. wartime low-tech, rather than adult "gossip"...) scheme whereby a set of physically separate loggers could sync and close gaps caused by local connectivity barf.

    It would probably be unwise to fully automate this, but even a semi-automatic sync would beat the current (afaik) state of the art (hands manually grinding through db dumps?)

  32. Mircea Popescu`s avatar
    32
    Mircea Popescu 
    Sunday, 4 August 2019

    They could just publish a clearsigned daily dump, similar to what phf publishes, and read it from each other.

    It'll keep it synced, it'll give us concrete data re risks of machine signatures, it's strictly ideal.

    And it is immensely wise to automate it, because the potential benefit is so large (someone burns a 0day ? really ?), and the potential loss so small (someone gets me to briefly appear to have said c0ks in january ? really ?!).

  33. Results of second run of btcbase snarf, for interested parties:
    http://blog.lobbesblog.com/2019/08/btcbaseorg-snarf-first-run/comment-page-1/#comment-41

    @Jacob Welsh Welcome. Re: Frantech. I would definitely read a blog post of your experiences with them; go for it! Though I have never given them a single bitcent (they only got my worthless fiats). I only went with them because they were cheap and, for a time at least, they served me adequately enough. No longer the case, sadly.

  34. @Mircea Popescu: do you have a position on the canonical internal representation of the log? (See e.g. http://logs.bvulpes.com/asciilifeform?d=2019-8-4#476072 )

  35. Meanwhile, phf came back to life.

  36. Mircea Popescu`s avatar
    36
    Mircea Popescu 
    Sunday, 4 August 2019

    @Stanislav Datskovskiy 1 The indexing of lines should definitely be per chan. The problem of how to maintain the loggers in line-sync is actually wider than that, though. I fear it may turn out that the naive hope is unsustainable, and we can't actually have portable line numbering across independent loggers ;/

    @Stanislav Datskovskiy 2 I actually agree with his position. His communication style is incompatible with any sort of position of social preminence of any kind, he's not "cut out" to be a lord of the republic in the same exact sense he's not cut out to be in charge of a three person daytrip.

    In particular brandishing about this fake dilemma of forcing people to either "wait quietly forever" or otherwise "turn out to have been such monsters later" is intolerable. I do not demand that anyone find time in their busy schedule of getting sepsis to inform me they're getting sepsis ; but I also do not have any intention of doing anything with anyone who will surprise me later. Inconvenient or however else, this'll have to be the absolute outer limit of things.

    For what it's worth, I do believe the fellow contains everything that's required to fix whatever's wrong with him. I wish him the best of luck in doing so.

  37. @Mircea Popescu / 1: If the line numbering is made per-chan (pretty simply, as I now realize, all that's needed is to turn it from "primary key" into ordinary integer, and instead primary-key on internal row #, we're querying 'where channel=...' anyway) why wouldn't the numeration then be portable? Would have to decide on a numeration start for the other chans, however. (From what begins e.g. #asciilifeform ? Ben never put up a raw dump IIRC.)

    @Mircea Popescu / 2: Difficult, IMHO, for mortals to make the hard-guarantee of "no surprises!" But this is why I gave to Mircea Popescu e.g. FG pcb masks, Phuctor backend snapshot (probably oughta do again, it has changed since, when did we last do this..?) etc. And everything else of mine that might be of any conceivable use after I get sepsis, is published... (What runs -- genesis'd; what doesn't run, or otherwise unfit for battlefield -- snapshot-cum-hash in logs.)

  38. @Mircea Popescu / 1: Thinking about it further: if logger gives on request the initial serial # of a particular chan, then it does not matter whether two loggers agree in re absolute numeration. For so long as the # increments monotonically, loggers can still compare dumps and find missing lines.

  39. Mircea Popescu`s avatar
    39
    Mircea Popescu 
    Sunday, 4 August 2019

    #1 : Because, conceivably, freenode presents lines in different orders on different servers. There's no hard guarantee of order in the irc protocol itself (granted, disorder rarely occurs in practice, but I suspect mostly because nobody is seriously using irc for any purpose).

    #2 : Yes, impossible. Nevertheless, here's one sad fact from a long list of sad facts dredged up by the "kinetics -- disposes" side of this particular unfolding of phenomena : opting to keep "girl" outside of republic comes with that unfathomable coda of a cost, that one day when you get sepsis from an utterly improbable chain of events, she won't know to say "hey guyse, my man is very sick" to thereby pre-empt the fallout from unrelated and equally improbable chain of events resulting in unpleasantness. If one botheres to calculate by-minute ROI of his women, I doubt that one particular minute comes out so low, it actually loses to ten thousand others ; but if indeed it does, then in all fairness why persevere getting in each other's way ? She didn't know she should have called, or answer calls, or SMS, or anything, at all. Fine then. What more ?

    In the most general form, the problems of socialization are not to be approached through technological means ; there are excellent solutions already available, tried and tested for many many more centuries than "technology" has had years. It's folly to eschew all that.

    And, of course, as you point out, the other sad fact from the pile : opting to not publish (even to the very modest standard of "publication" contemplated in the Republic) also comes with unfathomable codas, and so ongoing, throughout the list of sad facts. It is one thing to be insulted by comparison to Framedgragger ; it is quite another to look at actual history so as to see whether and what exactly the difference was supposed to have been at X point in time. What, Framedragger wasn't a delightfully intelligent fellow who contributed well working, useful services he never published and who disappeared midstep ? Are we all personally exceptional forevermore and that's that, the last word on any examination ?

    There's no shortage of sadness in the pile, and none of it reduces to "man should have been made of iron". On the contrary, on the exact contrary.

  40. @Mircea Popescu 1.1 : Was comparing own local logs to Phf's today; sadly already found instances of just such reordering.

  41. Re: log displayer: I'ma have to rewrite mine. It is rubbish, currently.

  42. @Mircea Popescu:

    > There's nothing wrong with implicit-&, if you want to pre-process the nude natural language sentence hurr durr and qurr into hurr&durr&and&qurr.

    Indeed.

    @Jacob Welsh, Mircea Popescu:

    > Does the fact that & is the delimiter for URL GET parameters argue in favor of + as the conjunction operator for search?

    On the other hand, + is used in some boolean algebra texts to denote disjunction, which makes it all the more confusing. So rather, *?

  43. Mircea Popescu`s avatar
    43
    Mircea Popescu 
    Monday, 5 August 2019

    Fuck those texts, I'm not losing + to them ; and besides * is anti-intuitive for this purpose. Dot (.) would be the one available option, but it's too fucking tiny and not visible.

  44. Mircea Popescu`s avatar
    44
    Mircea Popescu 
    Monday, 5 August 2019

    @Stanislav Datskovskiy Speaking of things that are total rubbish, your blog not having any kind of recent comments thing is proving to be a royal pain in the ass these days.

  45. @Mircea Popescu : re: recent-comments device : I entirely agree, it is quite annoying omission (afaik, the ancient WP I cribbed from, did not yet have it.) It is on conveyor right behind install of the modernized select displayer (my www is still using the old, barbarous one.)

    In fact I'd rather even switch entirely to MP's WP. But first gotta move the site to new hoster... and hands presently more than full.

  46. Mircea Popescu`s avatar
    46
    Mircea Popescu 
    Monday, 5 August 2019

    There is that.

    Not to mention Pizarro doesn't even have more slots >D

  47. @Mircea Popescu et al:

    Fixed the schema per Mircea Popescu's #36 .

    And I did find a clean method of fixing the highlighting (e.g. http://logs.nosuchlabs.com/log-search?q=2014&chan=trilema displays correctly, links are unmutilated). IMHO the displaytron is ready for battlefield test.

    The program currently running on Dulap can be seen here: http://p.bvulpes.com/pastes/Rw9RB/?raw=true

    At this point I would like to invite people to abuse the search (and any other abusable component), see if can find other oddities.

    If Trinque is tuned in: plz post your patched CL backend, where working multichannel, and detection of "silent" disconnect works? I would like to plug it in tonight.

  48. Incidentally, apparently ye olde a111 bot is synced, running and voiced (but looks like everyone is waiting for new logger? and green light from Mircea Popescu?), at the time of this writing no one has spoken in #t.

    @Mircea Popescu if the #t is meant to be still unplugged ATM, might want to unplug deedbot, theoretically people can speak (and slightly complicate the sync job of new logtron when it stands up).

  49. Mircea Popescu`s avatar
    49
    Mircea Popescu 
    Monday, 5 August 2019

    Nice work on the select huh!

    Re search abuse, I tried http://logs.nosuchlabs.com/log-search?q=everyone%27s+invited+on+trilema+%26+other+blogs&chan=trilema with no results. Am I doing it wrong ?

    On the "theoretically people can speak" angle... sure they can. I just won't ask for whatever they say be reflected in any logs, is all.

    I left trinque a ping on his blog re the CL.

  50. @Stanislav Datskovskiy: looking good, misses the beloved from:nick doohickey tho.

  51. @pletzalcoatl: "from" and other "intelligent" searchisms are not in the proggy yet.

    @Mircea Popescu: simple answer to the riddle -- I haven't yet synced the thing with Phf's *current* (i.e. working, a111 is up) dump, you will notice that my log ends when his bot fell. Naturally I'ma sync it before revving up the backend.

  52. FTR, other things not yet made in this displaytron include: MP's idea for "checkerboard" colours; Phf's classical (and I personally quite like) backlinks system; quoted phrase search; pagination of search outputs; marking of bot outputs (as in Phf's system.)

    It also currently has a notion of time at variance with Phf's, which results in variant calendarization of log pages vs. his; but I suspect this is fixable by adjusting "zone".

  53. Mircea Popescu`s avatar
    53
    Mircea Popescu 
    Monday, 5 August 2019

    Oh I see, I was doing it wrong. I'm guessing 2015 items out of scope the other way, also.

    Could you set it so it echoes the search to the form (simply populate the name=q value= value with whatever's in the ?q= GET) -- this should be in the standards revision also I suppose.

    Is this genesis'd / published somewheres or not quite yet ?

  54. @Mircea Popescu:

    Phf's edible .txt dump (see also comment #3) begins with April 2016 (aka "dragon").

    Did notice that there is no echo of the search in the form; will fix.

    Would like to genesis ASAP, but ideally *together with compatible back-end* (to which perhaps *someone else* can then sit down and add search command, logline echo, and other direly missing essentials, I would quite like to have hands freed up for backlog. Will update, however, my copy to include all subsequent properly v-published improvements, as I read'em) and setup instructions.

    If you think it makes more sense to genesis separately from back-end, can do it then as soon as tonight.

  55. BTW if you actually want #13's item 2 (and I suspect this goes for 4, in same, also) would have to use a custom-written indexer, rather than the DB's (and will likely lose your Greeks, and ditto "Кто ты по жизни" and all other hieroglyphs. The current item however *does* eat hieroglyphs, as Phf's did.)

  56. Mircea Popescu`s avatar
    56
    Mircea Popescu 
    Tuesday, 6 August 2019

    I think you should genesis the working object (ie, front + back end) as soon as practicable. Though genesis is perhaps the incorrect word, what is it, rather an extension atop trinque's bot or ?

    I have no strong feelings on the matter, both 2 and 4 seemed like a good idea from an usage pov, but obviously usage merely proposes...

    I dunno I've ever actually searched for non-ascii chars tbh.

  57. @Mircea Popescu #56: Working logtron consists of two separate proggies:

    (A) Front-end (as sitting on Dulap and pictured in the paste, and largely subj of thread here), this displays a log from given DB to http://www.

    (B) Back-end. Speaks IRC, sits on chans, processes commands, echoes (if voiced) etc. IIRC Phf's even knew how to self-voice via deedbot. My draft of (B) is a modification of "logbot-multiple-channels-corrected" (by Trinque & Ben_vulpes) but it has same weakness as e.g. Pehbot (fleanode will occasionally "silent wedge" it. Trinque has what appears to be a working pill against this, but has not published src yet AFAIK.)

    Processing of commands, and all live/in-chan operation, live in (B). Currently I have no commands at all, not even auto-echo, T & B did not see it fit to include any in the genesis'd item. This will have to be written.

    (B) is IMHO best baked as a vpatch sat on logbot-multiple-channels-corrected. (A) can be made either standalone genesis (it'll ride on any back-end, so long as uses Phf-style format and my schema) or rolled into vpatch predicated on (B), for a complete kit. My initial notion was that ought to have genesis of *complete* kit, such that could be stood up immediately by noobs with minimal instruction. This however may have to wait on Trinque. Alternatively could genesis front-end (A) separately.

  58. And hm, your WP seems to auto-htmlize my spurious 'www' in text?

  59. Mircea Popescu`s avatar
    59
    Mircea Popescu 
    Tuesday, 6 August 2019

    Well, if it's lead by html :// it's assumed to be a link, so yes.

    Re the design, since there's no utility to A without B, why isn't then A a patch atop B ?

  60. @Mircea Popescu/59 : Funnily enough it was a naked 'www'. As for patch, I have no objection to genesising'em as complete kit. But then oughta wait for Trinque. IMHO the unpatched backend isn't of much use.

  61. Err, patching on Trinque, that is, rather than ab initio genesis. IMHO it is rather ugly to house two independent proggies under one patch, however.

  62. Mircea Popescu`s avatar
    62
    Mircea Popescu 
    Tuesday, 6 August 2019

    But I mean, in what sense are they independent ? Seems to me A can't meaningfully run without B, it's like saying you want to house the house and the drywall under different roofs.

  63. Mircea Popescu`s avatar
    63
    Mircea Popescu 
    Tuesday, 6 August 2019

    www
    http://www.
    http://www.dot
    http://www.dot.dot
    :www
    ://www
    ftp://www

  64. Mircea Popescu`s avatar
    64
    Mircea Popescu 
    Tuesday, 6 August 2019

    Basically, if you add a dot past the www it'll take it as an url. Nfi, I don't remember where this knob is even.

  65. Update: search box fix of #53; bot markings (a la Phf's).

  66. @Mircea Popescu/62: Aite, let's go with patch-on-Trinque&co.

    Meanwhile, added "most recent line time" display in chan select bar.

  67. Mircea Popescu`s avatar
    67
    Mircea Popescu 
    Tuesday, 6 August 2019

    Exciting times.

  68. Meanwhile, #50 fixed, "from"ism added, e.g.:

    http://logs.nosuchlabs.com/log-search?q=trb+from%3ABingoBoingo+from%3Aasciilifeform&chan=trilema

  69. BTW open call for util that'd eat a ZNC log dir and shit Phf-style .txt edible by eater.py. I don't know if will have time to write it with own hands this week, but it is fastest way of filling in the log archive for the non-#t chans.

  70. Mircea Popescu`s avatar
    70
    Mircea Popescu 
    Tuesday, 6 August 2019

    Lettuce see what lobbes, spyked, trinque et all say when they wake up.

  71. Update: fixed ugly mistake where users with '_' in name (guess who!) would not display in 'from:' query.

  72. Update: "last delta" now clickable; decorations.

  73. @Mircea Popescu

    > Sure, as long as you stick around to learn from it.

    Thanks, and makes sense. I'm glad to see the more seasoned hands have made such strides on the logger situation, and at this point am thinking to continue with mine at a slower pace, given these observations:

    It's Python 3, and indeed the unicode-by-default thing increases the burden of understanding the code, especially for the bot component as IRC is based on bytes. It might work for now, but since the immediate fire seems under control, I'd rather avoid the trap of "temporary" solutions and redo in a proper language. Or at least a less improper one.
    It doesn't use SQL but runs directly from a text file. I'm rather fond of this design: it's much more fits-in-head and potentially a performance benefit, but will need some improvements to be adequate here (indexing, atomic writes with respect to the frontend, multiple writer safety).

    @Stanislav Datskovskiy

    The main difference I noticed in comparing your schema and phf's raw logs with my format is that the first two don't distinguish the IRC message type: in particular NOTICE which as I understand is intended for bot messages. How do you distinguish them in the frontend -- operator must maintain a list of known bots?

    > IIRC Phf's even knew how to self-voice via deedbot.

    Could be, but I noticed:

    16:09:32 a111!~a111@unaffiliated/phf/bot/a111 joins #trilema
    16:09:32 MODE +v a111 on #trilema by ChanServ!ChanServ@services.

    From an implementation view, getting voice directly from chanserv seems attractive as it avoids further dependencies in what's otherwise a fairly simple bot (http client, deedbot IP, GPG...), though this may be short-sighted, e.g. other IRC networks might not have compatible services.

    > fleanode will occasionally "silent wedge" it.

    One way this can happen is when client is waiting on server and server loses TCP state, e.g. is rebooted and a FIN doesn't make it to client. For this case, IRC PINGs are effective, or Linux has a TCP keepalive (which I haven't tried, needs to be enabled in a couple places, per tcp(7) and socket(7) ).

  74. I am parsing the html logs to try to reconstruct text files for reference and to get some sort of before 2016 data. This seems not too hard, the the html is well formed. I am currently stuck on some details (decoding html escape sequences), I will report progress.

  75. @Stanislav Datskovskiy, @Mircea Popescu: IMO the full logger orchestra should be patched on top of the current Ircbot/Logbot Keccak V tree (as mirrored by Lobbes and myself). The btcbase patch viewer still contains the old SHA512 tree from what I see. The "logbot" side of the tree also contains a Python command router implemented by Lobbes, otherwise command support would have to be imported from the "Trilemabot" branch.

    Re. providing separate front-end/back-end patches: this may make sense if someone at some point decides to implement e.g. their own IRC backend, but if so, then I guess that someone could also make a V patch. I personally went the other way in the past, i.e. I provided a patch for s-xml for whomever wants to use it in their project, and then I copied that into the Feedbot tree later.

    Re. Ircbot reconnect patch: it would be more than welcome, I intend to use it for Feedbot myself.

    @Jacob Welsh:

    > From an implementation view, getting voice directly from chanserv seems attractive as it avoids further dependencies in what's otherwise a fairly simple bot (http client, deedbot IP, GPG...),

    Would depend on Deedbot anytime rather than on Fleanode's infrastructure, for the simple fact that Deedbot's maintainer is in the WoT. The unfortunate truth however is that Trinque's bot does depend on Fleanode for operations such as cloaking, but I suppose that'll go sometime in the future, as Deedbot replaces Fleanode's bots.

    > One way this can happen is when client is waiting on server and server loses TCP state, e.g. is rebooted and a FIN doesn't make it to client. For this case, IRC PINGs are effective, or Linux has a TCP keepalive (which I haven't tried, needs to be enabled in a couple places, per tcp(7) and socket(7) ).

    Pings are already implemented by Trinque's Ircbot as per the spec, and yet Fleanode still drops connections silently and apparently randomly. I've tried enabling the keepalive thing myself some years ago and it doesn't do anything -- TCP is brain-damaged and that's that.

  76. Mircea Popescu`s avatar
    76
    Mircea Popescu 
    Tuesday, 6 August 2019

    @Stanislav Datskovskiy

    > (guess who!)

    ascii_field ? :D

    @Jacob Welsh

    > the trap of "temporary" solutions

    Worked great for lobbes, keks.

    I dunno, traditional wisdom was, baselessly, that "do it anyway, you'll at the worst learn from it", but I put a cork in that until lobbes gets around to confirming it, once the experience settles sufficiently in the past for him to become visible.

    > directly from a text file

    flatfile db gains ~nothing over actual db except for brevity ; but there's nothing wrong with it in principle, people use it all the time. Are you relying on some kinda package or just hacking one by hand ?

    (Incidentally, looking into "does a flatfile db standard even exist in python or wtf are they doing, https://campus.datacamp.com/courses/importing-data-in-python-part-1/introduction-and-flat-files-1 at random, holy shit does python downstream sound like a wanna-be TMSR / obnoxious crappleheads cult. Someone puts up with these ninnies ?! Leaving aside how they neither have a standard "use this for flatfile db" nor perceive they long ago should have, just the stridently whiny, zit-covered tone should be enough to send everyone to the hills)

    > NOTICE which as I understand is intended for bot messages

    This ain't so.

    > From an implementation view, getting voice directly from chanserv seems attractive

    If I recall correctly, I put this in at botop's request, so he doesn't have to do the proper voicing "right now". That, of course, was long ago ; the salary for my tolerance turned out to be degradation of the environment, of course, τὰ γὰρ ὀψώνια τῆς ἁμαρτίας θάνατος. Needless to say I'll be taking it out once I log back in, and not putting such wires in the fuses in anymore, either. Irrespective of orcs' notions of
    "efficiency", the fuses actually serve a role in urban environments.

    > IRC PINGs are effective

    This sounds theoretically like it'd work, but then again the bot already does pinging and apparently not that effective.

    @ave1 Cool!

    @spyked

    > on top of the current Ircbot/Logbot Keccak V tree

    Yes.

    > btcbase patch viewer still contains the old SHA512 tree from what I see

    It's part and parcel of how to make friends and influence people, this. God knows worked splendidly for mod6, too, it's just...

    Say, alf, did you actually fall for the poisonous offering, ended up taking dead tree for real tree ? Or were you looking at the proper one all along ?

    > then I guess that someone could also make a V patch

    Precisely.

    The only way it makes sense is if someone decides to use an extant front-end (eg, me, along the lines of the thing I suggested to Diana Coman, re using MP-WP to frontend). But I don't presently know whether I actually even want to do this (because), meaning I can always just re-genesis a portion of a tree later, if indeed need be, rather than do it now "just in case".

  77. Mircea Popescu`s avatar
    77
    Mircea Popescu 
    Tuesday, 6 August 2019

    Btw spyked re http://thetarpit.org/posts/y06/098-hunchentoot-iv.html#fn8 third comment "what am I gonna do" portion : you could simply run MP-WP on commodity imperial hardware/software stacks, which is what I'm doing.

    I'm not saying you have to, nor am I saying I believe you're doing it wrong necessarily ; but consider, what are your criteria when hiring sherpa (visual aid) ?

    If the criteria is, "hire sherpa who actualy gets you there", so far one sherpa takes you there but fucks you in the ass while camping each night ; while the other sherpa takes you to roadside teahouse with great view of destination mountain.

    If the criteria is rather, "hire sherpa who doesn't abuse my holes", it makes no sense to hire anyone at all -- being by yourself is the best guarantee available your holes will remain unmolested, hiring any sherpas at all seems imprudent.

    I'm aware that yes indeed, like I was saying on Diana's blog, better smart, beautiful, rich and healthy than dumb, ugly, poor and sick ; it's so much better to deal with Republican people on Republican infrastructure and so on. I see it, for sure, it'd be so much better if you could hire sherpa that took you there and also sucked your cock every time you felt like it on the way over. There's no argument here at all, I'm with you, complete agreement. I'm not for a second proposing not to hire this sherpa, not at all. Just...

    Obviously it's up to every man (& woman, & child) to decide when the overarching consideration is "eh what the hell, not like it's made of soap, to wear out with use" and when... not. This is why I'm not saying you have to or that you're wrong. Just... seriously now, how am I going to leave you comments ? I like talking to you, and this is getting somewhat ridiculous.

    Meanwhile re http://thetarpit.org/posts/y06/099-tmsr-work-iv.html#selection-229.119-229.222 (does this work the same for you, "which makes me seriously think that I should take the first Lisp texts and transcribe and annotate them" ?) : I believe it is absolutely true and entirely the case lisp will need a man in charge. The monkeys that "invented" it may have left behind some still-usable shards of the microscope they hammered, smashed and bludgeoned every possible boulder on their merry way with ; but what's truly needed is an actual master of lisp to play the traditional role of "buck stops here" that mastery is all about.

    This is not a function of "if we decide to use it" ; I am coming to believe lisp needs someone in that vein for and by itself, irrespective of anyone's usage. It's easy enough to mock the confused retards of the indirection-layer-secreting type, but this mockery is only telling half the story. The other half would be to point out that there is nobody capable, willing, and ready to summarize any portion of lisp in any given size arbitrarily chosen. This is to say : a) in order for lisp to be a thing b) there must exist someone who c) holds the entire, complete and correct tree of all of it in his cache and d) can, upon request, either explain why a question is malformed, or else return an exact subset of the language as required. Meaningfully.

    It's not merely a "SELECT * FROM lisp WHERE post_status like "publish" and `post_content` LIKE '% depth%'" producing an implicitly-ordered list. It has to be "Explain what Trilema thinks on seafaring, in three hundred twelve words". The bolded items are metasyntactly relevant, ~anything~ goes there. Anything at all.

    Lisp needs a Grand Oracle, as the only possible solution to "the shitsoup ecosystem" ; and as the only possible fruition of alf's "fits in head" conceptualization (yes, this is all the man's thinking about when he says that : a leg of this elephant, he understands one portion of what's needed to construct the possibility of utility in a rational intellectual process).

    I don't think you're qualified to do that ; nor am I asking you to do that.

    I'd have asked phf to do it, but an alt-imaginary phf, one who spent the last year chewing through things like early Mocky, running me out of what to say, "go to Qatar, why not" "Hi from Qatar", "look into gfx stack" "oh hi this is gfx stack". That imaginary phf I'd have asked to oracle lisp for us, because life must be run until it hits a wall, that's what it's all about. The actual phf as he actually is, rather than as one could have imagined him, found a short little hedge two driveways over to splat himself against instead (hopefully not permanently, as ~everyone else has already said), so no, I'm not about to ask him, either.

    That leaves us without many people who even give enough of a shit about lisp to pester with all this ; but what can you do ? It's unavoidable, functional programming can't go away just because we're all morons who are easily overwhelmed by abstraction and "how does it all relate to me" all over ourselves.

    In the circumstances, it is my considered belief that all we can do until the hero arises is polish the arms, like good little doobies. So yes, collecting and adnotating the remnants of Antiquity may even spark a Renaissance, who knows. Worth doing in any case, enjoy.

  78. @Mircea Popescu:

    > you could simply run MP-WP on commodity imperial hardware/software stacks

    I've actually considered doing this, but none of the options are a solution to the *actual* problem, i.e. that there currently exists (to my knowledge, at least) no Lisp blogotron with comments, pingback etc. functionality. And this shortcoming comes to me as motivation to solve the problem rather than work around it.

    The easiest and most convenient thing that I could do would probably be to revive the old blog while I work on adding comments to the new one, but then I'd have articles scattered all over the place, and folks'd be all confused about where I actually write. The next best thing would be to migrate everything to MP-WP and (when a proper Lisp blog CMS is available) back, but then that would take more work than implementing comments on the current thing, so... wtf?

    Then I could abandon this Lisp WWW thing until we get a proper Lisp and a proper WWW stack, but then... I've had my fill of PHP and on the other hand I don't know what other shit lies in the current WWW server stack that would inform future development. So I might as well find out and also fix my own shit while on it, and use this experience to then do the right thing(tm).

    > Just... seriously now, how am I going to leave you comments ? I like talking to you, and this is getting somewhat ridiculous.

    I'm fully aware of that, and heck, I'm fully taking the responsibility of people potentially ignoring my writing until this gets fixed. If my estimation is right, this sad situation won't last longer than a couple of weeks.

  79. Mircea Popescu`s avatar
    79
    Mircea Popescu 
    Tuesday, 6 August 2019

    Meanwhile I've greatly expanded the comment you were working on ; I guess I shouldn't have done that.

    > but then that would take more work than implementing comments on the current thing

    I am willing to be that thisis not so ; and not even close. I am willing to bet this is not so by a degree of magnitude. Just my private bet on the sidelines, don't mind me.

    Two weeks aren't that far away in any case.

  80. @Mircea Popescu, spyked :

    > Needless to say I'll be taking it out once I log back in, and not putting such wires in the fuses in anymore, either. Irrespective of orcs' notions of "efficiency", the fuses actually serve a role in urban environments.

    Looks like there is not currently a bot (other than unpublished deedbot) that even knows how to GPG. So, no echoing logger in #t until someone bakes one? (The other potential headache, is that deedbot will turn into point of cascading failure, as I understand.) What's the practical win from "all bots must ask deedbot for voice" ? It won't do anything against hypothetical malfeasance of fleanode.

    > but then again the bot already does pinging and apparently not that effective

    Trinque appears to have solved the "silent wedge of fleanode", somehow. But not yet published precisely how.

    > Say, alf, did you actually fall for the poisonous offering, ended up taking dead tree for real tree ? Or were you looking at the proper one all along ?

    Actually even different, I was working from my local copy, i.e. from pehbot src. Which does not contain the python knobs, but wasn't intending to extend the orig bot in python, would much rather put the knobs in the CL.

    Naturally vpatch ought to go on the Keccakized tree.

    > The only way it makes sense is if someone decides to use an extant front-end

    Front end is much easier to make than the back (observe, mine is a couple of pages), so I expect there would be almost as many front-ends as there are loggers. Which is why I considered to genesis it separately.

    > Lisp needs a Grand Oracle, as the only possible solution to "the shitsoup ecosystem" ; and as the only possible fruition of alf's "fits in head" conceptualization (yes, this is all the man's thinking about when he says that : a leg of this elephant

    IMHO a proper doing of this requires the full orchestra, i.e. proper reversing of the Bolix followed by lifting of the "good parts" and replacement of Common Lisp with rationally-designed item. On rationally-designed iron. Which would then make sense to "Grand Oracle". It is my belief that CL-on-PC is a dead end, and a particularly dangerous one which has led a great many people to their intellectual (and, in Naggum's case, otherwise) death.

    There's a reason I put the work into the Bolix dig, and it isn't because I like the smell of old irons, I do not sit and reverse old arcade machines, Ataris, etc.

    It is an item I would like to get back to after rollout of Peh. (Do not presently know when will get this chance, apparently all hands, incl. mine, on deck for re-implementing same old logger for 9000th time? And it'll have to GPG with deedbot... and digest fleanode barfs. For 9000th time, Phf! why not publish yours? Whatever you think is wrong in it, could be fixed!)

    > So yes, collecting and adnotating the remnants of Antiquity may even spark a Renaissance, who knows.

    I've collected these for over decade now, FWIW. But IMHO collected enough "pieces of the microscope", there is quite enough to make new one. All that's in short supply, is time. At least on my end.

  81. Thinking about this further, IMHO deedbot is prime candidate for the inevitable rerun of this thread, should Trinque happen to get appendicitis or similar. I'm not aware of him having a spare Trinque in his back pocket, either.

    Seems to me a solid argument for "if you want your item to be infrastructure, genesis it".

  82. Dear lordship,

    As some of you know, I have recently made some big changes in my private sphere so as to gain mastery over my own life. This has been going well, and I've never been able seen things so clearly. I've given up the pot, and dropped my own negative ROI girl from my life. However, there are still many things I must do to take back control, and I find I simply cannot keep up with the very real responsibilities of being a lord in the public sphere.

    This does NOT mean I am going to vanish. When I met with Mocky, I told him that the republic is one of the few things in this world that has given my life meaning. As long as both it and myself continue to stand, I intend to keep serving it.

    Essentially, I need direction. I am too distracted by my private sphere responsibilities to keep up with making smart decisions on my own in my public sphere. I think I could better serve TMSR if I had a lord or lords making those decisions for me.

    In other words, I'd rather be a respected knight than a flailing lord.

    When I get home to my pgp key I will reissue this as a signed statement.

    May the Republic prevail over all matters.

    -lobbes

  83. Mircea Popescu`s avatar
    83
    Mircea Popescu 
    Tuesday, 6 August 2019

    @Stanislav Datskovskiy

    > Looks like there is not currently a bot (other than unpublished deedbot) that even knows how to GPG

    Does this ridiculous situation strike you as tenable ? In fucking 2019 ? Are we crypto-whatever or aren't we!

    What am I supposed to do, avert the real pressure by immaginary fiat ? That shit doesn't actually do anything, shorting the fuses isn't a fix, you understand me, getting MP to say "oh, bots can't rsa and that's ok" doesn't MAKE it ok.

    > What's the practical win from "all bots must ask deedbot for voice" ?

    That we have bots that can rsa, and botops who have experience, over years with running bots that can rsa.

    And that we find flaws in deedbot -- both as the extant item and as its own conceptual fleet in being.

    > Trinque appears to have solved the "silent wedge of fleanode", somehow.

    I am not aware he has, there's the occasional deedbot down moment as well. I'm not even certain it's solv-able, but it's high fucking time we find out already.

    > Front end is much easier to make than the back

    I suspect this is naivite, but we shall discover in time.

    > IMHO a proper doing of this requires the full orchestra, i.e. proper reversing of the Bolix
    > followed by lifting of the "good parts" and replacement of Common Lisp with rationally-designed item

    I can certainly see the logic.

    > It is my belief that CL-on-PC is a dead end

    This might even be true, sadly enough.

    The one scary thing is that the market for people to program the resulting monster may indeed be the fabled three hackers, if we're lucking enough and they don't all get born in Minsk one certain generation. It's one thing to derride the "accessibility" parading in software ; but the other end of that pendulum is... gnarly.

    > re-implementing same old logger for 9000th time

    This is the punishment for our sins -- until we figure out how to do things we just re-do the done things to death.

    > I've collected these for over decade now

    Doesn't mean I wouldn't read his.

    > deedbot is prime candidate for the inevitable rerun of this thread, should Trinque happen to get appendicitis or similar.

    This is not even inconceivable.

    > Seems to me a solid argument for "if you want your item to be infrastructure, genesis it".

    There's two classes of problems with this. The first is that well, deedbot also has wallets, the responsibility for which gives the op significant say in the management of the thing. The second is the commerical angle -- even though trinque's not doing much (that I can see) to advertise deedbot outside the walls, and thereby drive adoption (in the exact manner pizarro is contemplated to), nevertheless the item itself is classified as a commercial entreprise, entitling the entrepreneur to some protection from disloyal competition.

    All of these being probably ripe for benefitting immensely from frank discussion, but somehow we haven't gotten to it as of yet. (Yes, it was my intention, years ago, to use pizarro's putative success at gathering custom to encourage trinuque to replicate it ; but that was a different universe.)

  84. Mircea Popescu`s avatar
    84
    Mircea Popescu 
    Tuesday, 6 August 2019

    @lobbes

    > I've given up the pot

    Does this mean pot as in, weed, smoking cannabis ?

    > Essentially, I need direction.

    Well okay, but can you explain what with ?

    Maybe that'll help, you know, write an article about yourself. On the upside, I'll certainly read it ; on the downside I know that these things always seem, subjectively as well as to (retarded) nearby advisors "a terrible idea", but... I dunno, Nicole doesn't seem to have suffered any for being caled Miss Piggy, in spite of what the old woman wailchorus I plucked her from "thought".

    As it turns out, truth is very powerful -- for that matter it'd have been the "considered" opinion of ~any and all "nobrainer" morons that it'd be a bad idea for me to publish either this or the other article. And yet...

    Worth considering.

    > I think I could better serve TMSR if I had a lord or lords making those decisions for me.

    That can also work, why not. Pick one, talk to them. Diana for instance seems to be actively open / trying to build the infrastructure for just such work, maybe you can help along there. Alf's castle also seems to be moving of late, if that sounds better to you.

  85. @mircea_popescu

    > Does this mean pot as in, weed, smoking cannabis ?

    Haha, equivocation strikes again. Yes, I meant smoking cannabis. That old drug of complacency. I was a heavy pothead for the last decade. Haven't touched it in a month now and the clarity of mind I have now is like a drug in itself.

    > Well okay, but can you explain what with ?

    Specifically, I need help prioritizing my TMSR work, as well as prioritizing what to learn and when. Mentorship, I guess, is what I seek.

    > Maybe that'll help, you know, write an article about yourself.

    I agree. I will do just that once I get home from the saltmines.

    > As it turns out, truth is very powerful

    Amen. If there's one thing I learned from this place early on, it is that pretense is a mind killer.

    > That can also work, why not. Pick one, talk to them. Diana for instance seems to be actively open / trying to build the infrastructure for just such work, maybe you can help along there. Alf's castle also seems to be moving of late, if that sounds better to you.

    I would love to lend my hands to both of them, if they'd have me. Though right now it seems alf's work on logotronics will directly help Diana's work on outreach (which we desperately need, as BingoBoingo I'm sure can attest to)

  86. Mircea Popescu`s avatar
    86
    Mircea Popescu 
    Tuesday, 6 August 2019

    Curiosity struck, because this sprawling comment section seems pretty hefty. Aren't you curious how it scores with historical Trilema ? Is this a particularly large set of comments ?

    One

    SELECT tp.post_name, tp.post_title, SUM(LENGTH(comment_content) - LENGTH(REPLACE(comment_content, ' ', '')) + 1) AS count FROM MPWP_comments AS tc LEFT JOIN MPWP_posts AS tp ON tc.comment_post_ID = tp.ID WHERE tc.comment_approved=1 GROUP BY comment_post_ID ORDER BY count DESC

    and 1.5528 seconds later, here's a top 25 :

    1. Vina batranilor : 51347 words ;
    2. Concurs de proza arhiscurta. : 19791 words ;
    3. De fapt, de ce nu folositi voi linux ? : 18985 words ;
    4. [Artifexd] A better ircd [RFC] : 18620 words ;
    5. Platforma unui posibil partid conservator roman : 18206 words ;
    6. Concurs de injurat blogari, editia a II-a : 17325 words ;
    7. Responsabilitatea si Serban Huidu : 17268 words ;
    8. A complete theory of economics : 16476 words ;
    9. [G]ossipd design document : 16410 words ;
    10. Concurs de injurat blogari : 15843 words ;
    11. I think you drastically overestimate the military ... : 14012 words ;
    12. Wikileaks - DCVIII : 13518 words ;
    13. Despre-un baiat si-o fata. : 13301 words ;
    14. Procesul Ceausescu : 12659 words ;
    15. Cum ne scapam de Bucuresti ? : 11426 words ;
    16. Putoiul online redux : 11298 words ;
    17. Concurs de proza arhiscurta, editia II : 11254 words ;
    18. Un blestem ciudat. : 10876 words ;
    19. Victoria Romaniei : 10815 words ;
    20. Bani pe blog. : 10555 words ;
    21. Cupluri si relatii : 10545 words ;
    22. #trilema goes dark : 10518 words ;
    23. O recunoaste cineva pe troglodita asta ? : 10133 words ;
    24. So who's running the Courts circus ? : 9916 words ;
    25. So you think you know about scams, right ? : 9875

    Data accurate as of my comment #84 above. So now you know.

  87. Mircea Popescu`s avatar
    87
    Mircea Popescu 
    Tuesday, 6 August 2019

    @lobbes

    > the clarity of mind I have now is like a drug in itself.

    I'm glad that's working out for you, then!

    &godspeed.

  88. @Mircea Popescu/83:

    > That we have bots that can rsa, and botops who have experience, over years with running bots that can rsa. And that we find flaws in deedbot

    It isn't impossible, or even unspeakably difficult, to write proggy that can de-pgp etc. What is more concerning is that it will turn deedbot (subject to all of the usual fleanode weather, as well as the frailty of flesh of author, as it is unpublished) into central column of failure.

    > there's the occasional deedbot down moment as well

    Seems like it has stood back up every time since he patched? (Admittedly I have not combed the log specifically for 'how many sec. of down-time each time')

    > I suspect this is naivite, but we shall discover in time.

    My front-end is AFAIK feature-complete, with exception of backlinks, pagination, and quoted-phrase search. (If can think of what else missing, plz say.)

    > The one scary thing is that the market for people to program the resulting monster may indeed be the fabled three hackers

    I regard this kind of thing as a failure *by definition.* Properly-designed lispm ought to fit *entirely* inside a few hundred pages of printed matter. Just as, incidentally, CL would have, if it were not designed by tower-of-Babel committee.

    E.g. Scheme v4 standard doc (complete with Backus-Naur-ism etc) was about 50 pgs.

    > deedbot also has wallets ... the item itself is classified as a commercial entreprise

    I'm sympathetic to "commercial enterprise" angle, but also find distasteful idea of hanging *all* other infrastructure on a proprietary mechanism maintained by just one (however strong) Trinque. (Did you even ask him whether he's ready to become this sort of Atlas? So far looks like not even replied to yesterday's ping?)

    > All of these being probably ripe for benefitting immensely from frank discussion, but somehow we haven't gotten to it as of yet.

    Ideally IMHO would have a working #t in which to discuss. Though this here format apparently also works quite well. Let's discuss.

    > to use pizarro's putative success at gathering custom to encourage trinuque to replicate it ; but that was...

    This touches on my own albatross, i.e. the still-unresolved (and stalled, now that hands full of "pay for sins, write 9000th logger etc) q. of "in what to house the heathens". The MIPS experiment was IMHO a complete failure. And it is impractical to house'em (supposing they were to turn up when invited) in shared unix hosting, the very first hooligan will wreck entire assemblage and I've not found any attractive recipe for preventing it. This open problem IMHO merits active discussion, esp. before BB & I commit to a new set of irons (the q of "which irons" still standing, as of time of this writing. Was orig. thinking, "obviously, large AMD, we'll fill'em with MIPS sessions. but now what.)

  89. To pile on top of 83/88 -- seems to me to be a waste, to pour sweat into "how to weather fleanode and TCP barfs" when the obv. Right Thing is gossipd. (Which, depending on whether or not Trinque and Phf continue to sit on rather than disgorge at least sections of their unpublished works, may in fact be closer to hand than a 100% mature 9001st logtron-cum-GPG-auth..)

  90. @ave1 / 74: See Mircea Popescu's #10; there is not any actual need to grind up Phf's html barf to extract pre-2016 material -- MP has it already in edible form. (As does Phf -- who turns out is alive and possibly in near future will stand up again and lend his hands)

  91. Mircea Popescu`s avatar
    91
    Mircea Popescu 
    Tuesday, 6 August 2019

    > it will turn deedbot into central column of failure.

    No, properly speaking it'll turn the central column of failure that is deedbot from perma-failed when it comes to voicing bots (as it stands now, since bots don't even know to ask) to perhaps-sometimes-working. I mean, it's not that if the bots start sucking somewhat less I thereby lose the capacity to voice them, or anything. Right ?

    > Seems like it has stood back up every time since he patched?

    I dunno, maybe. It's hard to tell, consider a111 was "good for a year", presumably also "after phf patched". What can you do ?

    > feature-complete, with exception of

    Right.

    > a failure *by definition.*

    There is no absolute rule promising you computers. You understand this, do you ? We may, at any point in our blind march towards nothing, we may discover there's simply a missing step. Someone forgot to add the superficial iron, there's coincidentally no way to properly make computers go, who the hell even knows.

    I'm not saying this has to be so.

    > Did you even ask him

    I did not in so many words ask ; nor am I trying to squelch your line of inquiry here, I'm just waiting for the man to speak.

    > Ideally IMHO would have a working #t in which to discuss

    I can certainly see the angle, sure.

    In fairness, this emergency test of backup resources didn't come out all that bad, but yes, I'd also prefer the irc format myself.

    > This open problem IMHO merits active discussion, esp. before BB & I commit to a new set of irons

    Certainly.

    Did you end up deeming it unfeasible to continue with rockchips ?

    You could, you know, simply host amds and only sell proper, whole, adult boxes, there's no absolute rule you must provide for the $5 market. For instance, I was considering another box to do another Trilema point over there.

    But maybe this should wait for the irc forum re-embodying itself ?

    Re #89, wouldja simmer down for one minute ?

    Re #90 I took it that he relishes the exercise, rather than perceives the dire need. Was this not so ave1 ?

  92. @Mircea Popescu (strictly in re 91) :

    > Did you end up deeming it unfeasible to continue with rockchips ?

    They "conveniently" vanished from market (near as I can tell.) This was in fact one of the thorns in my arse that prompted the MIPS experiment, recall.

    > You could, you know, simply host amds and only sell proper, whole, adult boxes

    May in fact have to. Which promises to be painful, heathens are fixated on "low cost at all cost" to just about the total exclusion of everything else.

    > But maybe this should wait for the irc forum re-embodying itself ?

    Depends, may take a month for self-voicing bot to appear. Would rather not lose entire month in re Piz, if can be avoided. I fully understand MP's distaste for popping in to other castles, but I will ask him to consider popping into #piz at some point so that we can expand this thread, when time permits.

    At this point I'ma have to "simmer down", must attend to saecular crapola.

  93. Forgot to add, in 91-92 -- the lack of working preemptible-thread GNAT for ARM is also a sticking point; even if I were to unearth a crate of old-stock RK3328. AFAIK there has been no progress on this front, and my own hands are not free to attempt it. There is also the q. of whether ARM is worth the sweat: AFAIK RK was the one single blob-free-bootable example of this machine on entire market. And now "mysteriously" shortage...

    I have a complete (pending genesis, was in fact in the process of cleaning the barf up and packaging) kernel, gcc, universe, for MIPS. But where to get a MIPS?

  94. Mircea Popescu`s avatar
    94
    Mircea Popescu 
    Tuesday, 6 August 2019

    > heathens are fixated on "low cost at all cost" to just about the total exclusion of everything else.

    This is specious reasoning for the following unexpected reason : heathen women are also fixated on spawning short daugthers with no tits, to just about the total exclusion of everything else also! Nevertheless, because they have many daughters, and I can't use that many girls for anything...

    You're already stuck with a selective approach by the simple fact that you can't (nor do you want to) host every one of the web's however many million websites. In this context, saying "items of interest occur in concentrations of 1ppm" isn't particularly meaningful -- your sack only holds a few pounds and the ocean's many million millions.

    > consider popping into #piz

    Wait, wait, the thing specifically constructed on the premise of my not being involved ?! Come on!

    Why the hell should rsa processing take months, there's already an eucrypt published properly and everything! Yesterday you were sounding like you're gearing up for imminent delivery, da fuck happened.

  95. @Mircea Popescu /94 :

    Point taken re '1ppm daughters'.

    > specifically constructed on the premise of my not being involved

    Lol, since went from "not involved" to #1 buyer. And strategic brain (which I appreciate.)

    > you're gearing up for imminent delivery

    Of logger? This was prior to "must talk to deedbot." If speaking here of RSA, mine needs Keccak. (Which is what was on my front plate prior to a111 fell.)

    Currently Trinque's item uses traditional GPG. And presumably new log bots will also have to, for the time being. More items that will have to be retooled...

  96. Mircea Popescu`s avatar
    96
    Mircea Popescu 
    Tuesday, 6 August 2019

    > If speaking here of RSA, mine needs Keccak.

    http://ossasepia.com/2017/12/07/introducing-eucrypt/ you know ?

    > Currently Trinque's item uses traditional GPG

    All deedbot wants is a decrypted string. I guess one might need some glue to go from the plaintext paste it spits out to rsa values, but I can't imagine it'd take a month.

  97. @Mircea Popescu /96:

    "decrypted string" of GPG sucks in massive rfc2440 spittoon of AES, SHA, etc. none of which are in Eucrypt (or FFA) or worth carrying into future IMHO.

    So not quite so simple.

  98. I have to say I'm pleasantly surprised the backup communications channels are working as well as they are.

    @Stanislav the problem with MIPS is a lot like the problem with ARM, where to get one that isn't blobbed "new MIPS" Octeon. On that front there's a number of Huawei and other routers. As far as higher density alternatives to 1U adult AMD 64 boxes, it doesn't look like there's anything other than the PCEngines APU on the horizon until we start printing our own chips. 4 APUs to the rack unit is still better density than the pilot RK plant. It still wouldn't be a thing for the 5 buck market, What it'd be is a capable, standardized web/irc/etc server for those not wanting to pay a whole rack unit for the extra compute horsepower.

  99. Mircea Popescu`s avatar
    99
    Mircea Popescu 
    Tuesday, 6 August 2019

    @Stanislav Datskovskiy This is a point.

    So basicaly the idea here is that it'd b best to get trinque to spit out ~two~ pastebins, one in traditional GPG and the other in republican format ? Then bots can use later and humans can in their own time transfer from past to future ?

    @BingoBoingo That's an idea.

  100. @Mircea Popescu, @BingoBoingo / 99 :

    Indeed , what I would like to do, ideally, is to wrap up the constant-time Keccak and then immediately go to Trinque to help him set up parallel (for time being) auth mechanism, using working Peh.

    Re: BingoBoingo's APU : the one concern is that they are somewhat poor price to horsepower ratio, and require custom metalwork (there is no heat sink mount as-such on the board, it has to be bolted upside-down onto a thermal pad.) But it may be the best available solution at the moment.

    Re MIPS, I'm quite aware that there is *no* usable commercial MIPS on the market today; have looked for one for years. If we go for MIPS, it will have to be a FPGA (to eventual ASIC) grind.

  101. @Stanislav Where the APU wins is not price to horsepower, but in it being the apparent AMD64 hygiene winner while still more importantly being a standard.

    It isn't unlikely that I'll get a motor vehicle down here at some point. That motor vehicle however will almost certainly be a rear engined, air cooled VW. Not because it offers the best price/horsepower, but because it's a standard, well documented thing with excellent parts availability. It's a tool for transportation, and 40-ish hp can do that.

    Similarly there's a lot of folks that want a hosted box that runs a bunch of relatively lean tools while connecting to the network. This is what the RockChip is though the Rockchip is handicapped on the ADA front, (to my knowledge) Cuntoo front, and the maker's commercial environment (mass market hygiene insensitive) front.

  102. @BingoBoingo / 101 :

    Unfortunately, the APU1 isn't exactly your "40hp VW" in re "excellent parts availability" -- the AMD G-series was proclaimed "end of life" several years ago by AMD, and threatens to go out of print any day (to be replaced with the Fritzchipped variant.) But your overall point stands.

  103. @BingoBoingo et al:

    Thinking further, the "oops, your cpu went out of print" thing will, I suspect, become a perennial plague when using anything other than konsoomer x86; already ate one such bullet with RK.

    This is specifically why I took an interest in MIPS, as it is the simplest field-proven design with *existing* compiler support, and entirely practical to implement quickly in physical iron if we get the chance. (The MIPS-on-PC emulator that I wrote, is actually a considerably more complicated affair than to bake actual MIPS! -- at least on the logical complexity, rather than manufacturing cost, end. Actual mips fits in

  104. grr, comment cut off on less-than sign;

    Actual mips fits in _less than_ 100 ln of Verilog, and IIRC illustrated this in #t on at least 1 occasion.)

  105. @Mircea Popescu in re orig subj. of this thread (logtron) : presently it seems that I am waiting for Trinque to wake up? Let me know when you decide that this wait has timed out, at that point I will stand up (and genesis) back-end based on old-fashioned, (presently wedgeable...) CL item.

  106. Err, vpatch, rather than genesis. (Apol. for the comment spam, I am rather abusing this thread as a substitute for #t realtime.)

  107. Mircea Popescu`s avatar
    107
    Mircea Popescu 
    Wednesday, 7 August 2019

    @Stanislav Datskovskiy I guess I can see it like that ; God knows the bots stayed derpy for years, a few more weeks won't make that much difference. So what's your eta on logger w/o rsa then ?

    > he "oops, your cpu went out of print" thing will, I suspect, become a perennial plague when using anything other than konsoomer x86

    I dunno, I'm happy with 20-year old amd CPUs to this day ; though these are extremely good value on all fronts (except they're kinda problematic for heat, which in a DC environment somewhat matters) they've been out of print for longer than Obama's been a thing.

    I don't see there's anything wrong with using the comment box to comment, so no apologies necessary ; I also can't imagine why you'd have to wait to vpatch. Stop inventing reasons to wait upon, vpatch all the things right nao.

  108. @Mircea Popescu / 107:

    > So what's your eta on logger w/o rsa then ?

    If I continue now (i.e. w/out waiting for Trinque) I expect minimal kit can stand up and run (with same modest degree of weather resistance as pehbot) before end of week.

    Then will need actual command processing (search, etc), right now there are none.

    > I dunno, I'm happy with 20-year old amd CPUs to this day

    For certain applications I'm even as happy with 40-year-old cpu... problem in re use w/ server is, in both cases, the bloat of "modern" sw (in particular 'lamp stack' & co.) Not to mention limited supply.

    Must also note, 32bit archs in particular are crippling in re large working sets (will just barely handle TRB, and can forget about large in-mem db etc.)

    > vpatch all the things right nao

    Will then vpatch backend on spyked & co's, entirely good point, no reason to wait.

  109. Mircea Popescu`s avatar
    109
    Mircea Popescu 
    Wednesday, 7 August 2019

    A bot without search and that keeps falling down is not actually much use.

    The amd fxs are 64 bit.

  110. @Mircea Popescu / 109:

    > not actually much use

    As can see with naked eye, the www end does eat search. Missing bit is in the irc-end command eater (I don't expect it will be esp. difficult, but must point out that not touched it yet.)

    > amd fxs are 64 bit

    Ha, hadn't noticed the link, thought contention was re literally 20y.o. cpu (e.g. the Athlon I had in 1999-2000 etc)

  111. Further in re 110+ -- I gotta ask Mircea Popescu to specify what 'bare min.' logtron ought to have, to be of practical use; so can determine what remains, of this, vs. current.

  112. Mircea Popescu`s avatar
    112
    Mircea Popescu 
    Wednesday, 7 August 2019

    Oh you meant search from the chan, I see.

    Bare minimum logotron must have v-tree, uptime so it doesn't keep losing lines, someone behind it with reasonable response rates to fix whatever occasional hickups, logline echo and usable web face.

  113. @Mircea Popescu / 112 :

    Do you reckon that there is anything essential presently lacking in the example (on Dulap) www end? If not -- I'ma proceed with the irc piece.

  114. Meanwhile, I went and wrote a back-end from ground... it needs a little polish though, before ready to get off test chan and into battlefield.

  115. @Stanislav Datskovskiy/90

    That's great. I'll publish what I have so far found and stop at that point. Currenlty 2 features of log break reverse engineering of the originals;

    1. The new way of quoting links with [][text] is sometimes not distinguishable from the old if given [][] plus it will add "http://btcbase/log" if the link was relative so now the inputs [2019-05-01.html][may] has the same output as [http://btcbase/log/2019-05-01.html][may]. So the original cannot be retreived
    2. The link parsing is not always right, for example in http://btcbase.org/log/2016-12-29#1592569 the original link was not parsed correctly which can be seen from the target line. I've not found a correction method for this one yet.

    The code to convert a logfile from html to text:

    http://ave1.org/tarpit/log/scrape_log.py

    http://ave1.org/tarpit/log/scrape_all.py

  116. Mircea Popescu`s avatar
    116
    Mircea Popescu 
    Wednesday, 7 August 2019

    Ftr I tried

    http://logs.nosuchlabs.com/log-search?q=f%3Amp_+%22ascii%22&chan=trilema
    http://logs.nosuchlabs.com/log-search?q=from%3Amp_+%22ascii%22&chan=trilema
    http://logs.nosuchlabs.com/log-search?q=from%3Amp_+ascii&chan=trilema
    http://logs.nosuchlabs.com/log-search?q=f%3Aascii+mp_&chan=trilema
    http://logs.nosuchlabs.com/log-search?q=from%3Aascii+mp_&chan=trilema
    http://logs.nosuchlabs.com/log-search?q=from%3Aascii+mp&chan=trilema

    all of which produced 0 results.

  117. Oh ha, partial username q's. I'ma make these go.

  118. Mircea Popescu / 116:

    Update: all the given examples other than the ones featuring double-quotes now work. (Will make it eat double-quote, later this wk)

  119. Mircea Popescu`s avatar
    119
    Mircea Popescu 
    Wednesday, 7 August 2019

    Nice! Confirmed to work.

    Btw, I much prefer the f: style, so if there's any cost to also offering the from: style I see no problem in taking it out.

  120. @Mircea Popescu / 119:

    No measurable cost; I'ma leave'em both for now.

  121. Mircea Popescu : in case wasn't clear from the proggy -- is sewn in such a way that the older search links in log, still work (e.g. if not specified chan param, defaults to trilema, and so on.) Hence 'f' and 'from' and so on.

  122. BTW Mircea Popescu: would like to req command prefix for bot, and ideally two identical barf signs (e.g. two pounds, two carrots, two ats, whichever.)

  123. Mircea Popescu`s avatar
    123
    Mircea Popescu 
    Wednesday, 7 August 2019

    #121 A okies then!

    #122 Well, deedbot has !!, which is the only double sign seeing how they start with !. Looking through the list, the next available's !X, how about that.

    I guess one of the things to do after we get the irc going again is clearing out the dead wood in there, and in general, huh. Anyways.

  124. @Mircea Popescu / 123 :

    Will propose !N then, given how it's 'nsabot'.

  125. Mircea Popescu`s avatar
    125
    Mircea Popescu 
    Wednesday, 7 August 2019

    Alrighty, you can have !N, whaynot.

  126. Mircea Popescu`s avatar
    126
    Mircea Popescu 
    Wednesday, 7 August 2019

    Meh, on second thought it's kinda distant and prone to !B/!M misspelling... wouldn't you rather have I dunno, !1 ? It's right there...

  127. !1 works.

  128. Meanwhile, http://bvt-trace.net/2019/08/fg-fed-linux-rng-work-schedule/

  129. @Stanislav Datskovskiy, @Mircea Popescu: !X is IIRC taken by auctionbot and !1 was stolen by feedbot last year. So I'm thinking !2 or !q (as per the "harder to miss key" thumb rule)?

    @lobbes: Could pl0x update the list to reflect the changes?

  130. Mircea Popescu`s avatar
    130
    Mircea Popescu 
    Wednesday, 7 August 2019

    Awe shit, the list was out of date even ?! Nuts.

    Well Stan... what's it gonna be, !2 or !q ?

  131. @spyked Good call. I'll update teh list tonight once home.

  132. why not !q .

  133. Mircea Popescu`s avatar
    133
    Mircea Popescu 
    Wednesday, 7 August 2019

    Aite, !q it is!

  134. And list updated: http://lobbesblog.com/static/tmsr_bot_directory.html

  135. Mircea Popescu`s avatar
    135
    Mircea Popescu 
    Thursday, 8 August 2019

    You left out feedbot.

  136. Ah whoops. Fixed, ty.

  137. Update: bot is ready for battlefield (self-reconnects, correctly cycles through servers, etc.) ; tomorrow will stand it up.

    Still need that znc-eater to fill gaps in archive. Lobbes stated in #a that he picked it up. If tomorrow he doesn't turn up with it, I'ma write it myself.

  138. In other finds, apparently orcograms in search do not currently work. Will fix tomorrow.

  139. Mircea Popescu`s avatar
    139
    Mircea Popescu 
    Thursday, 8 August 2019

    Cool deal.

  140. @Mircea Popescu :

    One item I forgot about earlier: do we want link transformation? (e.g. when click btcbase from snsa logtron, go to the latter rather than former) Or only for 'era 1' (b-a, when we feed these in) ?

  141. Mircea Popescu`s avatar
    141
    Mircea Popescu 
    Thursday, 8 August 2019

    I don't think we do, no. People can use whatever links and more generally words they want. If not what's next, replace obama with bahamas ?

    I don't necessarily think we want them retrospectively, either : the bot readouts are there, so nothing's lost even if it should go away ; and why alter history ?

    I might change them on Trilema eventually, but that's rather a different story altogether.

  142. @Stanislav Datskovskiy re: znc-eater: Saltmines have been eating a lot of my time this week, but tonight I will be clear to get some more work done on that item. I'll keep you updated in #a once I get back home tonight.

  143. @Mircea Popescu :

    Thought was, would use same algo as Phf did (no "altered history", i.e. old links *print* same as they appear in a naked txt log, but the *clickable link* goes to the logtron that is displaying them, rather than to the historic one.)

    When we add in the "era 1" logs, will defo. have to do what Phf did, as the old #b-a logs site has been extensively monkeyed with by Kako.

    Q was whether or not we want similar link sugar for civilized/living bots. As I understand, answer is nope. So I'ma leave it alone for now. We can put in later if needed.

  144. Mircea Popescu`s avatar
    144
    Mircea Popescu 
    Thursday, 8 August 2019

    Yeah, I see what you mean ; but yeah, can be later, there's no pressing rush and let's distinguish the urgent from the cosmetic.

  145. Update: experimental draft bot is standing and logging (in #a, #t, #p, #e, #o) . Search and echo appear to work correctly. If no one finds surprises, will clean & genesis as it stands.

  146. hmm, already found one oddity : it wouldn't join #t (may be connected to fleanode retardation where they demand ssl/sasl in +v chans ??)

  147. Indeed,

    :hitchcock.freenode.net 477 snsabot #trilema :Cannot join channel (+r) - you need to be identified with services

    Dafuq, we're ident'd. Just not with SSListic services.

  148. Mircea Popescu`s avatar
    148
    Mircea Popescu 
    Thursday, 8 August 2019

    This kinda makes no sense, seeing how I never ssl-identified as far as I can recall.

  149. @Mircea Popescu : I entirely agree that it makes no sense. Yet bot is regged, auths w/ pw, sits down without complaint everywhere else but #t. Any idea for what to probe ?

  150. @Mircea Popescu : found fix. apparently gotta pause for fleanode to actually eat the auth, before joining.

  151. @Mircea Popescu : bot is now live in all config's chans, incl. #t.

    Please abuse in #a, #p, where low-traffics, and then time to genesis!!

  152. Also, closed the gap in #t log, using Phf's (live) raw dump for August. I.e. #t log as displayed is AFAIK current.

  153. Mircea Popescu`s avatar
    153
    Mircea Popescu 
    Thursday, 8 August 2019

    Anyway, so far no word from Trinque, BV or substantially phf.

    It seems safe enough to assume scriba is never getting fixed, so I'm just going to strike it from the list.

    I'm also going to take phf's lateral & oblique comments / silences as a simple statement that he is actually fucked in the head, and in pointed and deliberate (if elaborately disavowed) disdain of my expressly stated wishes his only real commitment is and remains this transparently ridiculous if outright imbecillic game of "force the others to act on no data only to whine at them later about how misunderstood he was for systematically keeping mum when he was supposed to speak, while deliberately organizing complex systems geared towards justifying that failure rather than preventing its occurence".

    Needless to say, I am unamused ; and, to answer the original inquiry in firmer terms containing no ifs or buts : no, I personally have no further interest in hearing what phf may have to say on any topic. The time for "ok then, I will get my logger to spec by X date and hope to have my blog up by Y date" came and went, sometime yesterday. The time to whine about "what it all means" etcetera is henceforth, 'till forever.

    I don't see the problem with any willing lord accreting this sorry carcass along with all the other hanger-on "unmotivated", misunderstood, mistreated & abused nonpareils -- though in all frankness I expect he's going to turn out the exact same sort of wasteful timesink as all the other sorry sacks presently trying to leech the life out of the remaining lords while posturing & pretending variously.

    I expect this will actually be a major problem for any and all castles going forward : all the fuckwits who tried and failed to dampen trilema will move on to trying to dampen whatever other castles, and perhaps not fail. While I'm firmly committed to the principle that each lord should run his own castle by himself, the unnatural situation whereby scum simply floats downstream endlessly retrying its scummy attachment strategies poses some significant dangers that require some reflection. I can't help feeling in some sense responsible for the burden, and so let me very firmly state my committment to making whatever I can available to the lords working to mitigate it. At the very least I will answer any question you come up with in your effort to run your castle, frankly and as quickly as possible.

    I don't imagine ben vulpes has any interest in doing any work ; so I'm going to also strike his logger from the list.

    From what I understand, lobbes was originally going to make a logger as per spec (by mid August or so), but then renounced that idea as far fetched.

    This then leaves yours to be on the autovoice list (as per #107), and perhaps spyked's, which was supposed to be maybe a September delivery iirc. I'm guessing that if any actual loggers show up by year's end I'm going to keep the logging thing open, and if not I'm just going to create a Stan monopoly on logging and be done with trying to push sheep up this particular hill.

    That said, 6d 13h after blackout (by NSA clock), #trilema looks ready to resume service. Ima log in half hour or so and see to the admin tasks.

    Congrats to the winners, as the expression goes.

  154. Noted and thank you.

  155. @Mircea Popescu I suspect Phf, fwiw, is still in bed. So I would not hurry to bury him.

  156. Mircea Popescu`s avatar
    156
    Mircea Popescu 
    Thursday, 8 August 2019

    Not a matter of burying. Fellow has more important / interesting stuff to do, and I have no intention of getting in his way.

  157. Aaand genesis posted.

  158. Mircea Popescu`s avatar
    158
    Mircea Popescu 
    Friday, 9 August 2019

    Add link plox!

  159. Genesis in http://www.loper-os.org/?p=3452 .

  160. Mircea Popescu`s avatar
    160
    Mircea Popescu 
    Friday, 9 August 2019

    Sweet.

  161. Meanwhile: raw log export knob.

  162. @spyked

    > Pings are already implemented by Trinque's Ircbot as per the spec, and yet Fleanode still drops connections silently and apparently randomly.

    I see; from a look at the code this appears so. (What's the spec you mean, as perhaps I've missed it?) So theory is that the server continues sending timely PONGs but stops forwarding channel messages? Wow.

    @Mircea Popescu

    > Are you relying on some kinda package or just hacking one by hand ?

    By hand.

    > https://campus.datacamp.com/courses/importing-data-in-python-part-1/introduction-and-flat-files-1 at random, holy shit does python downstream sound like a wanna-be TMSR / obnoxious crappleheads cult.

    Page doesn't load for me besides a husk and same in archive.is so I'll take your word for it. Maybe it's why GvR stepped down...

    > they neither have a standard "use this for flatfile db" nor perceive they long ago should have,

    If they did I'm not sure I'd trust it. They seem to care more about whether the code is "beautiful," in the eyes of parenthesis haters, than whether correct. (For example https://bugs.python.org/issue12268 still incompletely fixed in 2.7, whereby I/O calls can fail, and with data loss apparently, in the presence of signal handlers: another manifestation of the exact situation discussed as a Unix fail by Gabriel, "The Rise of Worse is Better", 1991.)

    > This ain't so.

    Thanks for the correction, looks like I oversimplified in my head on first read of the RFC section.

    > τὰ γὰρ ὀψώνια τῆς ἁμαρτίας θάνατος

    Speaking of the wages of sin, I'd better confess to Greek illiteracy (and I suppose this must mean illiteracy in general). Do you have any choice reference works to recommend?

  163. Mircea Popescu`s avatar
    163
    Mircea Popescu 
    Tuesday, 13 August 2019

    Page doesn't load for me besides a husk and same in archive.is so I'll take your word for it. Maybe it's why GvR stepped down...

    Item read

    datacamp-logo
    Course Outline
    Exercise
    Exercise
    Why we like flat files and the Zen of Python

    In PythonLand, there are currently hundreds of Python Enhancement Proposals, commonly referred to as PEPs. PEP8, for example, is a standard style guide for Python, written by our sensei Guido van Rossum himself. It is the basis for how we here at DataCamp ask our instructors to style their code. Another one of my favorites is PEP20, commonly called the Zen of Python. Its abstract is as follows:

        Long time Pythoneer Tim Peters succinctly channels the BDFL's guiding principles for Python's design into 20 aphorisms, only 19 of which have been written down.

    If you don't know what the acronym BDFL stands for, I suggest that you look here. You can print the Zen of Python in your shell by typing import this into it! You're going to do this now and the 5th aphorism (line) will say something of particular interest.

    The question you need to answer is: what is the 5th aphorism of the Zen of Python?
    Instructions
    50 XP
    Possible Answers

        Flat is better than nested.
        Flat files are essential for data science.
        The world is representable as a flat file.
        Flatness is in the eye of the beholder.

        Take Hint (-15 XP)

    Hint

    Note: If this is already loaded, executing import this will not print anything to the shell so you'll need to reload it using the following code chunk:

    import imp
    imp.reload(this)

    Did you find this hint helpful?
    YesNo

        IPython Shell
        Slides

    To me it reads exactly like Joseph Smith.

    Do you have any choice reference works to recommend?

    Perseus ; though mind that the item in question is more important in connotation than denotation -- my expectation writing it, for instance, was that the mind of the reader will summon the voice of Burton reading the English version in the 1967 film (not a good film ; but an excellent voice). This stands for most of classical text : not so much what it was, but who did what with it.

  164. @Jacob Welsh / 162 :

    While it may be early to say for certain, so far my bot has not seen a single dropped connection (other than when deliberately reset by my own hands.)

    So currently I suspect that the various historic workarounds (e.g. proactive client-side PINGism, etc) might be doing more harm than good.

  165. Mircea Popescu`s avatar
    165
    Mircea Popescu 
    Friday, 16 August 2019

    Aug 16 12:55:05 mp_en_viaje let's do some selected testing then.
    Aug 16 12:55:06 mp_en_viaje U+00B5 µ 0181 µ Micro sign
    Aug 16 12:55:18 mp_en_viaje U+00D8 Ø 0216 Ø Latin Capital letter O with stroke
    Aug 16 12:55:36 mp_en_viaje U+0126 Ħ 294 Ħ Latin Capital Letter H with stroke
    Aug 16 12:55:48 mp_en_viaje U+0193
    Aug 16 12:55:48 mp_en_viaje Ɠ
    Aug 16 12:55:48 mp_en_viaje 403 Latin Capital Letter G with hook
    Aug 16 12:56:05 mp_en_viaje keks even my term chokes on that shit.
    Aug 16 12:56:09 mp_en_viaje U+0193 Ɠ 403 Latin Capital Letter G with hook
    Aug 16 12:56:23 mp_en_viaje U+01DA
    Aug 16 12:56:23 mp_en_viaje ǚ
    Aug 16 12:56:23 mp_en_viaje 474 Latin Small Letter U with diaeresis and caron
    Aug 16 12:56:33 mp_en_viaje U+01DA ǚ 474 Latin Small Letter U with diaeresis and caron
    Aug 16 12:56:49 mp_en_viaje U+0211 ȑ 529 Latin Small Letter R with double grave
    Aug 16 12:57:05 mp_en_viaje U+0236 ȶ 566 Latin Small Letter T with curl
    Aug 16 12:57:27 mp_en_viaje U+03D6 ϖ 982 Greek Pi Symbol
    Aug 16 12:57:55 mp_en_viaje U+0426 Ц Cyrillic Capital Letter Tse
    Aug 16 12:58:11 mp_en_viaje U+0469 ѩ Cyrillic Small Letter Iotified Little Yus
    Aug 16 12:58:30 mp_en_viaje U+049C Ҝ Cyrillic Capital Letter Ka with vertical stroke
    Aug 16 12:59:29 mp_en_viaje

  1. [...] In response to: http://trilema.com/2019/trilema-goes-dark/#comment-130477 [...]

  2. [...] Trilema this morning (as promised, the signed version proving it was from yours truly), and it was Mircea Popescu's prodding to write about myself that prompted me to create this post. So here it [...]

  3. [...] the mice... Yes, I cribbed these last eight words from the man himself [↩]they say the accent varies by region, but fuck it I grew up there and I can't tell the [...]

  4. [...] order to keep track of and reference conversations in my #ossasepia channel. But on top of this and as mentioned previously, providing a public logger for tmsr channels is a basic contribution and one where redundancy is key in the long term. Not to mention that running a bot as part of a [...]

  5. [...] a comment box on this here Tar Pit; comments on the quality of Hunchentoot code; and an emergency halt to discussions on #trilema, caused by btcbase logbot downtime and the lack of an immediate [...]

  6. [...] even be retrofitted into something that fits in head, as it's otherwise by all looks as usable as a CL-on-Unix-on-top-of-TCP application can be. It's not like there's a better CL item to work with anyway, so [...]

  7. [...] like more than halfway through this series and most of what we've discussed comprises TCPisms and CL-on-PC. So let's begin our incursion into the subject with a [...]

  8. [...] next Hunchentoot demo will have to materialize into a minimally-working WWW-facing logotron. As I see it, writing a piece of software [...]

  9. [...] which was to be present and communicate clearly. [↩]I registered in the WoT in July to see #trilema go dark a couple weeks later, followed by Pizarro and now currently processing seemingly irreconcilable [...]

  10. [...] Stanislav Datskovskiy's ad-hoc python-based IRC logger bot he made for #trilema the irc channel during the previous crisis and the MP-WP LAMP-based web publishing mechanism supporting Trilema the blog. As I take it [...]

  11. [...] to be honest, I'm doing this with some reluctance, since I'm already going through some of the inherent pains of operating an IRC bot with [...]

Add your cents! »
    If this is your first comment, it will wait to be approved. This usually takes a few hours. Subsequent comments are not delayed.