The Freenode issue
38.229.70.22 aka card.freenode.net (via PSINet Inc, a name you really should know) disappeared early this morning. What's left is
;; QUESTION SECTION:
;irc.freenode.net. IN A;; ANSWER SECTION:
irc.freenode.net. 47 IN CNAME chat.freenode.net.
chat.freenode.net. 47 IN A 139.162.227.51 << aka niven.freenode.net, usg (via linode)
chat.freenode.net. 47 IN A 195.154.200.232 << aka barjavel.freenode.net, usg (via online sas)
chat.freenode.net. 47 IN A 185.30.166.38 << swedish mom&pop isp. not responding hencei.
chat.freenode.net. 47 IN A 185.30.166.37 << same swedish mom&pop isp. also not responding.
chat.freenode.net. 47 IN A 104.200.152.162 << usg via their LA cyberderpitude unit, aka "Total Server Solutions L.L.C." (the dorks forgot to add the PTR Resource Record, herpderp.freenode.net or whatever it was "Private Internet Access" allocated for them.)
They've been probing it for weeks now, and it seems to me likely there never will be a non-usgtronic freenode server henceforth. Thus the question before the forum becomes, do we wanna move off freenode now ? Or should I start peering the usgtards now and move later ?
———- Recent sampler, for the curious. [↩]
Tuesday, 12 March 2019
Are these peerable without a nod from the other end? Or is it like that SKS thing, where you gotta sweet-talk one of the existing derps into plugging you in?
Tuesday, 12 March 2019
They're peerable for me in the sense that they can't practically deny it.
It's just... by the time the whole thing's maggot by mass, what does sticking with it say about one ?
Tuesday, 12 March 2019
Peering might be interesting from an entomological / "nobus observatory" POV, but probably simpler to simply bake own irc relay set.
Tuesday, 12 March 2019
The simplest would actually be to join an extant irc network, one that isn't as "stupid" as Freenode-in-the-process-of-being-usg-infiltrated was, and will therefore federate our own servers. Recall, I offered them this back in 2012, the treasonous morons wouldn't hear of it.
Grow a server base (as well as the experienced operators capable of running it) in that shadow, while passively benefitting from the ddos-dissipation it naturally offers, and then at whatever later point we feel like it / the circumstances merit, separate our own network.
Freenode is intellectually dead anyways, is the overriding concern here -- consider the thick, rich line uniting those morons with those other morons... it's kanzures all the way through, what. It's both unbecoming and dangerous to one's mental health, wallowing in retard swamps. Who knows, maybe some other network has actually thinking people there. I expect (on the basis of experience!) that it is quite feasible to teach computing to ignorant thinking people -- god knows it is impossible to teach thinking to computer-cvasiliterate morons.
Tuesday, 12 March 2019
> extant irc network, one that isn't as "stupid"
IIRC this dig turned up ~empty last time we looked. (EFnet?)
Tuesday, 12 March 2019
I don't see anything wrong with trying efnet tbh.
As per netsplit.de, the top five largest irc networks are IRCnet (29k users), Undernet (18k users), EFnet (17k users), QuakeNet (14k users, who play or played quake, which seems no deterrent to me), Rizon (13k users). Freenode doesn't even figure on the list, and is in any case negligible.
Honestly, one idea would be to produce a multi-network bridge, have our channels in ~all~ these networks.
Tuesday, 12 March 2019
Speaking of netsplit, it would seem it reports a bunch of other active Freenode servers (e.g. wilhelm and hitchcock) that aren't found in chat.freenode's A records.
Tuesday, 12 March 2019
Well, it's a webpage, I expect they're synchonous to some sort of Ms standard, rather than s.
Tuesday, 12 March 2019
I like EFNet; it's a network I already connect to. There are a couple problems I can see: there is a smaller nickname character limit (my freenode nickname is too long, and so yours must also be), and there is no NickServe or ChanServe (meaning anyone can squat on your nickname and keep it from you if you stay disconnected long enough to let them).
I think both of these problems are not real problems for the Republic if deedbot is reconfigured to not care about what nickname you are talking to it from. It could get confusing for logbots and users in general.
There's another network I connect to called DALnet. It's similar to freenode in that it has NickServ and allows for longer nicknames. Although their nickserv requires that you finish registration via email and you must do so from the same IP address that started the process.
Tuesday, 12 March 2019
@Mircea Popescu et al:
As I understand, the various extant ircd's aren't actually interoperable in their peering ends. This AFAIK is the reason why there isn't an existing multi-net bridge. Currently if you want to peer with $net, you're stuck using (or emulating) their particular brand of nonstandard "embrace & extend" of the original '80s irc protocol.
Tuesday, 12 March 2019
@Daniel P. Barron
> I think both of these problems are not real problems for the Republic
They're not, in the sense we can just configure our ircds to not permit any of the bullcrap.
@Stanislav Datskovskiy This is so. The idea was to :
1. Pick one (I suppose, one of the top 3), add a server to their network, let the lord in charge of it have a little time to figure things out.
2. Pick another one (of the same set, or top 10, or w/e), add a server to their network also, run from the same box as 1).
3. Mutualize 1 and 2, by creating a wrapper layer that crosstalks, so that users connected to the box in question appear connected to our channels in ~both~ networks (and add courtesy trims like dpb discussed above, where our box also does not permit squatting of our names, by eg creating fake users to sit there silently until the actual person shows up).
4. Extend 3, one other network at a time, until the republic figures in ~all of them.
Tuesday, 12 March 2019
Most likely I'm being thick here, but why burn time creating a multi-network bridge rather than just settling for a tmsr-only-network?
If it is just for outreach purposes, then we'd still have the logs, blogs, etc for that. If user wants to talk real-time then they can connect to the theoretical 'irc.tmsr.net' and be done with it.
Either way, I'm for moving off freenode, but I agree: the q's are "where" and "how"
Tuesday, 12 March 2019
@Mircea Popescu:
The "wrapper" thing can be done, but at the obvious cost of the hypothetical relay box having the combined "porosity" of all N various heathen ircd's.
Tuesday, 12 March 2019
@lobbes Three reasons : to talk to people ; to avoid our enemies ; to learn how to do things.
What you call "burn time" is time spent on research for the future gossipd. The only thing you can't burn are development stages. Time can only be burned.
The "we already have X" approach is fundamentally broken -- why talk to girls on the street, we already have fetlife profile ? Because girls on the street, that's why the fuck talk to them, what!
Finally, if we make a three server concoction labeled "Hit here to disrupt Republic" we'd (for the first time yet!) do thinks stupidly. Our servers belong among the others.
In any case, we'd connect to 'irc.80.90.100.110" or w/e the box in question is -- the rest is supposed to be done by it.
@Stanislav Datskovskiy This is so, and I rather deem it a good thing. If nothing else, it's the sort of comment#3 research that's actually worth doing.
Tuesday, 12 March 2019
@mircea_popescu Ah this makes sense to me now. The "hit these three servers to disrupt Republic" point specifically is what made this obvious to me. Thank you.
Time to for me to dive into educating myself moar on the IRC protocol.
Tuesday, 12 March 2019
Come to think of it, the multi-network bridge approach even answers correctly the "where" question: everywhere. The concern is at most "where to start from."
Tuesday, 12 March 2019
Basically it'd be guided by would-be operators' personal preference alongside the constraint of which of the extant networks are actually federable in the first place.
Wednesday, 13 March 2019
@Diana Coman, Mircea Popescu:
Could even start with ye olde fleanode, at that. (Given as there is already Trinque's published routines for talking to fleanode.)
Wednesday, 13 March 2019
Coudln't, because Freenode will not peer with your ircd. That's the meaning of "federation" as it recurs above, those networks that will let you contribute a box.
Wednesday, 13 March 2019
@Mircea Popescu et al:
One, afaik, still open question: how to do user (and relay peer) auth?
The classic irc offers two just about equally rubbish choices -- sslism (what most of the nets currently use, and what fleanode increasingly forces on everyone), and plaintext salted hash (i.e. where it is ~trivial to hijack a session.) Both of these, imho, terrible. Do we pick one (sslism? where "nobus"? or salted hash, where "every douchebag can hijack"?) or have in mind a hypothetical third?
Wednesday, 13 March 2019
@Diana Coman:
I read comment #2 as "we could, even with fleanode, if wanted to." Am I mistaken?
Wednesday, 13 March 2019
@Mircea Popescu:
Pick one of the non-flea nets, that can be peered with, to start experiment with; I'll set up a relay to it on Dulap.
Wednesday, 13 March 2019
Certainly not sslism.
Considering we don't want to rewrite ~clients~ before we can sit down and have a civilised irc chat again, it's obvious we're left with the hash. Yes, it has many and plurious warts and disadvantages, and one overpowering advantage : it's what it is.
Which brings us directly to the discussion of
> trinque: the upstack claim was that this will help design gossipd. *only* as antipattern and study of antipatterns does not by itself yield sense.
It is false the effort will only produce data for study of antipatterns (while it is true that the study of antipatterns alone is not by itself liable to produce useful results). The effort will also produce data for ordering of future effort, which is not at all a negligible matter, even if it is traditionally neglected by engineers. As things stand, "write a gossipd" is too large a byte to take (in fact, as experience shows, even "write ffa chapters" can be on occasion too large bytes to take, and woe to the naive manager who says "tryna herd these cats is trivial and only results in antipatterns which aren't useful -- witness how fucking useful their dispassionate contemplation was to you!)
Which takes us to
> asciilifeform: so ftr i will have to disagree with hanbot , in that imho coal miner is ~not~ the best geologist.
This may be true, if what one reads is [expert] geologist [who is veteran miner also]. Nevertheless, coal miner has plenty of advantages over geologist fresh out of school (and let the disbeliever read Mark Twain, it's in there).
Unrelatedly, it's also true work in the coal mine narrows perspective, but this doesn't happen because of the work, but because of the exclusion of alternative. As an experimentally verified matter -- the brain of the woman who spends a few hours a day raising children stays in working order while the brain of the woman who spends no hours a week doing anything else BUT raising children turns to mush ; and so following with bureaucrats, and all others. I didn't learn communism through simple contact, either. See ?
Wednesday, 13 March 2019
As to comment #2 : I meant I could simply pretend nothing happened (like everyone else) and connect "as before" except through one of the usgistani servers. This wouldn't include any standing up of ircds, but merely standing up of a lot of sand to bury heads into.
I figured it's worth talking about ; the situation is actually eerily reminiscent of the sino attack on bitcoin a few years ago, we could say there are no independent peers left on the freenode network.
Wednesday, 13 March 2019
@Mircea Popescu:
Looks like all that's missing in the recipe then is "which net for start".
Wednesday, 13 March 2019
@Mircea Popescu:
In re: "usgistani servers" and "sino attack" -- I thought it was clear as early as '16 that fleanode was ~100% usg-owned. Hence why deedbot etc., rather than relying on nick auth for anything but decoratives.
But for so long as it reliably threw packets to where they're addressed (in addition to wherever else for the pleasure of the beetle-men) tolerated, because it isn't as if plaintext packets moving over the backbone aren't already straight for the beetlemen.
Wednesday, 13 March 2019
Alright, so we can have a day to pick one, I have not very strong an idea what'd be preferable but have sent off some people to come back with some answers, maybe that clarifies.
Re :
> trinque: ah, I hadn't even realized he couldn't connect *at all*
It's not that I can't connect at all -- I could connect right now, exactly like everyone else currently does. The whole issue is that I noticed this morning there's no remaining independent endpoints ; to connect I'd have to put one of the derps in hosts. In preference of doing that, I did this, which is even why the article's phrased as a question!
Also, I don't know how slow this thread will be, necessarily. So far the response time here is a few minutes as opposed to the irc's few seconds, but I daresay the time needed to think things through's still the largest draw, followed hopefully not too closely by the time needed to write, edit and redact one's thoughts in sensible form. Whether this happens one+ line at a time as in irc or one+ paragraph at a time as on blog should really make very little difference on the final time the whole thing takes, wouldn't you say ?
> But for so long as it reliably threw packets to where they're addressed
This is technically correct, but also the hard limit. Apparently there was a soft limit of "no independents left" that nobody knew about ("but mp, why didn;t anyone know about it ?!?!" "guess" etc). Seems the consensus is to ditch the fleanode anyways, so no real harm done here.
(Honestly, I was expecting an aparatchick showing up to scold us over our incorrect worshipping of the great mother taking poorly to ridicule with exactly the same effects, but apparently the almighty in their own mind are too weak and disorganised for anything like that.)
Wednesday, 13 March 2019
Fwiw, in re fleanode, 130.185.232.126 (bulgaria) seems to still answer.
( What was the litmus for "independent" fleanodes, anyway ? )
Wednesday, 13 March 2019
To clarify #7: I've tried both wilhelm.freenode.net (164.132.77.237, France?) and hitchcock.freenode.net (130.185.232.126, Bulgaria) and they both work. The list on netsplit.de seems to be updated pretty often (e.g. at one point there was also a hobana.freenode.net hosted in Romania, but it seems to be refusing connections now), no idea why chat.freenode.net DNS chooses to act retarded -- and if I understand correctly, nor that it matters much.
Regarding the broader point, I think
> Grow a server base (as well as the experienced operators capable of running it) in that shadow, while passively benefitting from the ddos-dissipation it naturally offers, and then at whatever later point we feel like it / the circumstances merit, separate our own network.
makes perfect sense.
I personally have no previous experience with standing up IRC servers, let alone with peering/bridging, but I'll start reading right away and I'd be happy to look into adding another server to whatever network we end up on.
Wednesday, 13 March 2019
@spyked The only interesting question here is, can you believe how much their DNS listing changed ? "Coincidentally", right. It just so happened that the whole structure, left undisturbed for as long as we've been using irc, suddenly got many and plurious updates within the hour. It's just how the world works (tm).