A collection of distinctions

Saturday, 03 December, Year 8 d.Tr. | Author: Mircea Popescu

Effort to understand the world around may result in discussion or conviction. It is generally a bad sign (for the survival fitness of the individual engaging in it and the group that tolerates the behaviour) when the result is convictioni, for which reason it's ever encouraging to see the Republican tradition of discussion flourish (within the well designed, and well protected, walls of the forum).

As part of the general effort, alf produced a piece (Vectored Signatures, or the Elements of a Possible V-Algebra) which, as he correctly points out, should be read. I'm not convinced it is the definitive word on the topic, or even definitive form of a point therein involved, but one item definitely stands out, undefeated :

Not only are there times when one would like to seal a payload with a caveat of one kind or another, but presently we have no means of conveying disapproval – other than by refraining from sealing. The latter act conveys very little useful information, and no permanent sealed record remains of the effort taken to actually understand the patch. This is a Bad Thing.

The entire collection of human scienceii is made up of exactly this : stories of what didn't work, and propositions as to why. Certainly for the infantile mind, still groggy under the alcoholic fogs of an artificial environment where a faux positivism of the nature of Santa Claus and Unicorns reigns supreme it would appear otherwise, but that confusion quickly disappears once the child starts doing real work. Once he joins the ranks of the thinking adults (generally through adding his own bug reports to the immense pile carefully collected over the generations intricately detailing the universal brokenness all around) he sets aside the ways of childhood and all is well.

This said, and without prejudice to the parts I either don't understand or don't have an opinion on, let's go into a detailed discussion of a certain aspect of the proposed scheme.

If one has a bit, one has two states, and if one had two bits, then one'd have four states. This much, being math, is not open to surprises. If one has a machine of the ilk of what today's humans use for computing, one probably has a notion of byte, which is wholly arbitrary yet experimentally supported - we're now joining science to our philosophy. On this shaky, chimeric basis, we can proceed to say that... we shall have four categories, and four states for each. This makes up... eight bytes, so it does something towards satisfying the practical angle. It also is symmetrical, so it does something towards satisfying the ideal angle, also. Maybe it's even a good idea! Is it ? Does it satisfy the right parts of each domain, does it take the rigurously correct steps ? This is after all what system design is all about : to first create the list of possible steps, to then sort them, and to then cut through the list at defensible points.

Maybe it does, I can't readily evaluate this question. How would we evaluate it ?

Alf has found labels for the four categories (hands, eyes, brain, heart), which are neatly symbolic and seem to work as a meaningful set. Is this proof that the correct cut has been made ? The statement is not without merit, a very practical understanding of "language as a mechanism for thought - benefitting even the dumbest members of a population from some measure of the collected intelligence of the very best accreted through the ages" would propose that the simple fact that you can find pleasing labels to create such a naturally satisfying set is an indication of correctness.

Maybe there's merit to this notion, maybe there's a reason we don't deeply, intuitively feel that nfgrl is missing, and that really any enumeration of hands eyes brain heart is incomplete without nfgrl. Is it happenstance we're born with five fingers ? Is it the case that the four organs Egyptians identified mean something a lot deeper than words can state, is it not mere coincidence the classic antiquity thought in terms of four humours ?

Fine, so maybe it's right that they should be four, and these four. Yet, wouldn't you find it peculiariii if it were the case that all four had equally reasonable, well spaced out and obvious - four degrees ? I would. I find it highly suspect, but then again what can you do.

Let's look at them, which is why the title reads "A collection of distinctions" : I will endeavour to define the edges between alf's states, because what could it possibly hurt ? What follows below is a best-effort attempt, and more as an exercise than anything, dedicated to the vague notion that even if the result isn't guaranteed to be useful or valuable, the path there certainly is.

1. The difference between "I did not create or modify any part" and "I did create or modified some part" is certain and immediately evident, of the visceral strength of "have you fucked your sister".

2. The difference between "I created/modified some part." and "I created most, or modified the original beyond recognition." is certainly weaker. For one thing : rooting a box is this procedure whereby through modifying "some part", one modifies the whole thing. The cultural practice (as distinct from the economic activity) actually recognises and rewards disparity for its own sake - the smaller the modification that modifies the whole entirely, the more respectable. What then does it mean, where computers are concerned, to distinguish between these ?

But perhaps the distinction has nothing to do with computers and everything to do with the humans supposedly riding atop them : the first is a change of implementation, details, we removed the spleen so infection won't spread and one leg of the bridge so boats can pass, but otherwise it's still a human and still a bridge versus we replaced the spleen with a live triceratops and we tore down the bridge and built a tunnel under the river, so it's no longer a human and no longer a bridge but still serves the original function, perhaps as redefined.

If this is the case, then the two should read "I made modifications of the implementation, preserving the original design" versus "I made modifications of the original design". Or maybe there's more cases available, in which case... I don't know.

3. The difference between "I created most, or modified the original beyond recognition" and "I claim sole authorship." is entirely of a different substance than the other two. What would it mean, in this context, to "claim sole authorship" ? Against whomiv is this claim levied, what effects is it contemplated to produce ? Is it a nude statement of fact, exactly equivalent with "I was wearing my shorts when I wrote this", and if so, why was it preferred over evidently available alternatives ?

Perhaps a much stronger and also much more usefully meaningful "I claim no alternative designs may be found that can resolve this problem" would be preferable ? To indicate that the author has presumably mathematical proof of the situation ? The only problem is that if such a situation arises, we would need a fifth state : there's a substantial difference between "this is provably the only available design" and "this is provably the only implementation of the given design". While often conflated in common parlance, the two are substantially different, and clearly at least as different as at least one of the previous distinctions.

5v. The difference between "I read none. Made no attempt to." and "I read some part; and/or skimmed some or all." is very dubious, because the first statement fundamentally invalidates the entire construction. If the author has read nothing, then why would anyone care what he signed, or how ? "Let me tell you what I think about things I don't know", really ?

6. The difference between "I read some part; and/or skimmed some or all." and "I read most of it." is altogether not clear. I for instance routinely read most of pieces through the procedure of skimming some or all. That's the whole point of skimming, and if your skimming doesn't result in this you're doing it wrong.

7. The difference between "I read most of it." and "I read all." is a flattering ideal, but sadly without practical basis. Try and name an item that you have read "all" of. Really, have you ? Unless the item is very short, and very banal, complete lecture is an unattainable goal. Consider this text :

Aaaaaaaaaaanaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaoaaaaa
aaaaaaaaaaaaaaaaaa

Have you read it all ? If you have, then whichth element is the o and how many a's are there ? Oh, you've read it all yet can't answer any questions about this thing you read ? This does not bode well, and while I'm aware that if you were going to sign this update you would have sat down and counted, nevertheless be aware that complexity quickly diverges from this three lines three symbols example.

9. The difference between "I do not understand any." and "I understand some, and/or poorly." suffers not only from the problem discussed in 5 above, but also from a subtler if related issue : how do you yourself know the difference between "I understand nothing of this" and "I understand some of this, poorly" ? It is paradoxically quite likely for the one who evaluates in the first manner to actually have a better understanding than the one who evaluates in the second manner (for a discussion of this see eg DK or commentary thereof). I have no idea how this distinction is supposed to be used by thinking people.

10. The difference between "I understand some, and/or poorly." and "I understand." has the same issue as the 2nd part of 9 above ; and as 7 before it. I myself understand many things, and my understanding of them is quite serviceable to me in my daily life. Nevertheless they're all fragmentary, and poorly understood on any sort of evaluation.

The race of understanding is ran with other people, not with the Concepts themselvesvi, and for this reason any serious attempt at evaluating this distinction reduces to a computation of the understanding of all others, which is unfeasible and for this reason shouldn't be introduced.

Moreover, why wouldn't one who understands poorly make inquiries in the forum, to remedy the situation ? It seems substituting that with signing is the wrong reaction.

11. The difference between "I understand." and "I understand absolutely." is perhaps not even as much of a problem as the others, because it could cheaply be translated into something meaningful, I expect on the lines of auctoritas : "I understand how this works but I wouldn't dare change it" versus "I understand this to the degree that I can fully produce the list of effects of any contemplated change, as well as propose alterations to any contemplated change so as to selectively shape the identified effects". This is that old, fundamental distinction between man and bureaucrat, wherein the former has the authority to change his mind, whereas the latter has the shackle of policy and consequently no need for a mind at all. In common software parlance this bit would make the difference between "I identify myself as a contributor to this item" and "I identify myself as a maintainer of this item", which is not a distinction without merit - if for no other reason then because it stems from actual practice.

13. The difference between "I distrust absolutely." and "I distrust." runs into a conceptual problem, in that a negative can not be absolute. This is no small matter, you can't distrust absolutely anymore than you can make an absolutely empty container. You can, of course, trust absolutely, or love, or seek absolute beauty - but there isn't such a thing as absolute hate, nor for that matter absolute ugliness. The absence of trust will have to stay relative as such.

Obviously a different term could be used, to make the distinction "I loathe" vs "I distrust" more or less reflecting the thought difference between "I am certain this will end poorly" and "I am not convinced this will end well", which isn't at all meaningless or uninteresting.

14. The difference between "I distrust." and "I trust." is not problematic, even if it could be discussed at length and colored variously.

15. The difference between "I trust." and "I trust absolutely." perhaps could reference the difference between what's called an educated guess and actual proved certainty, as in "I trust the RSA is safe" versus "I absolutely trust 7 is a prime number". Understood as such it is not problematic, even if it may be readily abused.

There are left aside some interesting bits, however. What about an item which breaks backwards compatibility ? Shouldn't the reviewer be able to make a note that the item reviewed breaks compatibility, perhaps with a distinction between "with another item I actually use" / "with another item I don't use but think should be preserved" ? This is a contextual consideration, of course, but certainly a major point to be considered when reviewing the utility of any communication scheme, seeing how most disputes in software management to date centered around an issue of deprecation / backwards compatibility.

What about items which are optimized for specified cases ? Consider the discussion of the way grep works. Presumably in a V world a common base of that project would at some point have received that patch. To some, such a patch is an improvement ; to others, it is breakage. Am I necessarily required to vote and pick one ? What if I agree that in most cases the sell-out is a good idea, however I would also like to preserve the other variant for the convenience of people who run specific tasks that would benefit from it ? How do I sign "this branch is good for X but not intended for Y" ?

How do I sign something as simple as "This man is a traitor" to contend with nonsense like Koch & friends constantly pump into their trees ? What, are we to imagine that as their fiat empire sinks into irrelevancy they'll just quietly keel over and die, perhaps on occasion agitating on their forgotten venues about nonsense ? No, they'll drown us in contributions, how do I nuke them ?

———
  1. At what point are you convinced ?

    This is no small matter. Consider : you go to the tap, lift or turn it, and water flows. But are you convinced it will ? Suppose it does flow - are you convinced it is actually water ? Yes, you may act as if you were, much in the manner a child too old may pretend to fall for the parental tricks not because he's mentally retarded, but because he has a big heart. But are you convinced ?

    You were, before I asked, right ? Which is the correct functioning of this mechanism, it should work if unchallenged and collapse with challenge. Yet this is not conviction, is it ?

    Conviction can't come from matter ; it is never nor can ever be some sort of byproduct of a structured, intensive, vast or otherwise process upon the world. Conviction isn't some sort of distilled reason just like airflight is not some sort of intensively trained self-wedgie application.

    Conviction is an abstract result, it comes from consideration of ideal objects not found in nature, and for this reason the conceit of deriving conviction from "facts" and "reasons" betrays a very broken mind and nothing more. []

  2. Math isn't science, for the needs of this piece, but philosophy. []
  3. These considerations, by the way, are very close to the game designer who no doubt lives in all of us. How shall you call the things in your game, and why ? And how many ? Of each ? []
  4. Yes, all claims have to be brought against a definite something. []
  5. You see what I did there, do you ? []
  6. Two men are surprised by a lion and take to flight. The one falling behind inquires with the one running ahead why bother, as the lion can easily outrun either of them ; and receives the answer that the forerunner aims not to outrun the lion - but his fellow man. []
Comments feed : RSS 2.0. Leave your own comment below, or send a trackback.

12 Responses

  1. FWIW I collected the equivalence classes first, and concocted labels -- second.

    Let's take, for instance, the "hands", and from my POV, some items I have signed in the recent past:

    Hands(asciilifeform, gnupg-1.4.10.tar.gz) == 0 ('I did not create/modify any part.')
    Hands(asciilifeform, tinyscheme.tar.gz) == 0 ('I did not create/modify any part.')

    Hands(asciilifeform, trb) == 1 ('I created/modified some part.')
    Hands(asciilifeform, sane-mpi.tar.gz) == 1 ('I created/modified some part.')

    Hands(asciilifeform, $certain_commercial_programs) == 2 ('I created most, or modified the original beyond recognition. ')

    Hands(asciilifeform, v.py) == 3 ('I claim sole authorship.')
    Hands(asciilifeform, this comment) == 3 ('I claim sole authorship.')

    These equiv. classes are not necessarily aptly titled, nor all-encompassing (e.g., how should Zapruder sign his film of the Kennedy assassination, without claiming authorship of the corpse, and/or of the bullet, at the same time as the film?) but they cover four very distinct situations that I, in particular, regularly find myself in when signing.

    The linked essay contained a handful of other examples.

    And the "must be able to sign without endorsing -- and in fact to convey the very opposite", and likewise to sign without implying that one has gone over every bit of the payload with a magnifying glass and understood every conceivable implication thereof - was the impetus for the whole thing.

  2. The above, continued, pointwise:

    1) 2, 3) claims of: none; or insignificant; or majority; or total authorship -- involve very different levels of ultimate responsibility taken for the payload. Declaring that 'if you touched it, you are an author just as if you had written the entire thing" may stroke your philosophical fancy but it is precisely the kind of thing that keeps folx who ought to be signing v-patches from doing so. I will not be taking responsibility for, e.g., Satoshi's rancid effluvia simply on account of having written several lines of patch for same.

    5, 6) '...If the author has read nothing, then why would anyone care what he signed, or how ?...' -- is broadly wrong. Both you and I have signed documents to the effect of "x is the tarball I downloaded in 2011 from y, and no further claims are made."

    Likewise I can usefully sign output from an RNG, without having pumped it through hexdump and meaninglessly into own eyes first.

    Consider my recent signature of mod6's 'parachute'. It merits a '1' in 'eyes', and very certainly not a 3, nor a 2.

    7) The chosen example is a poor one: if it were 100kB long, you would by no stretch of the imagination refer to it as 'human-readable' text -- any more than 100TB of cat /dev/random | hexdump -C are.

    9, 10, 11) Didn't this post contain a "I did not understand the thing in its entirety..." somewhere..!

    13) See the endnotes 10...13 to the linked post. The "trust" category attempts to convey whether the payload contains deliberate misrepresentations of fact, expressed or implied, and ought to be regarded as an "anti-endorsement" rather than the opposite.

    > How do I sign something as simple as "This man is a traitor"

    This is what 'signed WOT' is, as I see it, for!
    As per recent #t thread.
    Incidentally, is there any particular reason why a signed personal WOT could not coexist with the traditional type ?

  3. One could simply link a deed in the rating's notes field if he wanted to make statements of the "I swear this man is a traitor" variety.

  4. Mircea Popescu`s avatar
    4
    Mircea Popescu 
    Tuesday, 6 December 2016

    @Stanislav Datskovskiy Other than the general habit of not addressing the discussion, but unrelated generalities, specific problems are :

    the kind of thing that keeps folx who ought to be signing v-patches from doing so

    I do not care, nor does it make a difference, what "keep"s folk who "should be" eating from doing so. If they do not eat they will die, just exactly the same as the folk who walk in front of traffic, drink every murky liquid they run across etc.

    There's this misplaced perception of optionality, as if V is here to serve and support the needs of programmers as they perceive them. Nothing could be further from the truth - V is here to prevent programmers from doing most of the things they do today, completely and permanently.

    I will not be taking responsibility for, e.g., Satoshi's rancid effluvia simply on account of having written several lines of patch for same.

    If you make a patch, you sign that patch. Did we get confused as to how this all works and are now operating under the momentarily delusional notion that signing an element signs the whole tree or something ?

    Both you and I have signed documents to the effect of "x is the tarball I downloaded in 2011 from y, and no further claims are made."

    This is a temporary situation due to the sad realities of bootstrapping. I do not wish support for it be included in any sort of protocol, because I do not wish to see anymore of it. Going forward, none of this bs.

    Likewise I can usefully sign output from an RNG, without having pumped it through hexdump and meaninglessly into own eyes first.

    No idea what sense this makes. To "usefully" sign and publish output from a RNG ?

    To reiterate : the whole thing isn't about what you could do, but rather on the contrary - about what you don't get to do. V isn't here to defend your rights as a programmer. It is here to defend ~the world~ from the idiocy of programmers.

    Think of it like this - we're one short step from rounding up everyone who even touched a keyboard of any kind, shooting them in the gut and leaving them to bleed out. The one thing that may, perhaps, allow survival of computer programmers is V.

    It's altogether unclear it will work, in the sense that even if it gets universally and disciplinedly adopted there still remain plenty of futures in which all programmers are still rounded up, shot and left to bleed. Nevertheless, it's the only hope you've got.

    The chosen example is a poor one

    Actually, the reason you give for it being a bad example is exactly what makes it a great example.

    Didn't this post contain a "I did not understand the thing in its entirety..." somewhere..!

    For which reason it's neither signed nor in the V tree.

    why a signed personal WOT could not coexist with the traditional type ?

    In general it is a bad idea for multiple solutions for the same problem to "coexist" in this sense - much like it's a bad idea for Windows to "coexist" with software.

    Anyway, I suppose I misspoke, should have said "this is trachery" rather than "this man is a traitor".

    @Michael Trinque The entire point of this discussion is machine-interpreted signatures, ie the (tenuous, to my eyes) argument that something could be gained by allowing the computer to obtain more information than the simple yes/no was it or wasn't it signed.

    I very much doubt this is worth doing, chiefly because it removes human agency at a juncture computers are notoriously horrible. Let humans be gatekeepers, and if you want your software included get a powerful man to sign off on it ; rather than empower a dubiously shambling horror to make automated design decisions on the software I'll run.

    Basically - again - V is here to clip the exceedingly abundent wings of code monkeys, not to give them more shit to fuck up.

  5. @Mircea Popescu :

    As I understand it, you don't particularly care what the possible sig equivalence classes are, seeing exactly one safely-usable class, "read this human-readable message" ? Manual reactor control ?

    This is defensible but it puts a great deal of pressure on the human-readable caveat component, which is also how meatworld legal contracts ended up weighing what they weigh, 100s of KB... And they still get talmuded over. Is it possible to zap the talmudism, or at least reduce it to manageable proportions, in favour of a small and unambiguous command set that covers common cases ? And in your view, "I witnessed this", "I read this", "I wrote this" are not meaningfully-distinct cases ?

    And do you have your own, entirely different answer to the "how do I sign a formal fatwa against payload P" puzzle? I'd like to hear it.

    Elsewhere: Does the machine ~have~ to be a "dubiously shambling horror" ? And supposing that it were, then why would the dubiousness not extend to your RSAtron, your display handler, and other safety-critical components ? This argument is reminiscent of 1970s airplane pilots' objection to "fly by wire" - "what if the circuit fails?" To which the inevitable counter is "what if the wings fall off."

    As for "programmers", they are doomed, and they know it (this would handily account for some of the mental diseases seen in that profession...) , no one will save them. Talmudism leads to the lampshade, as drink leads to cirrhosis. If there is a long-term future for "computer", it is as a very minimal thing, that does not require "programmers" in the customary sense, only a literate operator.

  6. Mircea Popescu`s avatar
    6
    Mircea Popescu 
    Tuesday, 6 December 2016

    As I understand it, you don't particularly care what the possible sig equivalence classes are, seeing exactly one safely-usable class, "read this human-readable message" ? Manual reactor control ?

    Not necessarily. There are two distinct problems here.

    Problem #1 is that V does not exist so as to allow maximum expressivity for the coder, "here is one of the numerous internal states my head could find itself in, how do I use V to best express this". It is not crypto-powered livejournal for geeks ; the patches are not myspace status updates and so on.

    On the contrary, V exists as a harness on the coder, "here's the list of acceptable internal states your head may find itself in, work butt off until this ideal is realised in practice". It guides the activity of coders so as they serve society, the general good, the republic or how else you'd name it ; rather than the customary nonsense now current where "let the kids do whatever they want and we'll deal with the fallout later" - customary and current nonsense that nevertheless amply proved its unsustainability.

    Global warming is not an issue, even if the loud barely-literates claim it to be ; global conceptual fuckification however is, even if the same barely-literates pretend to not notice it.

    Problem #2 is that V is a tool for software (and system) design. In the way this currently works, with boolean signatures, the computer makes ~no decision~. All decisions are made by humans, their results being predictable and trivially enumerable.

    In the way a vectored signature V would necessarily work, the computer will make ~at least some~ design decisions. The results of decisions made by humans become significantly less predictable, and utterly un-enumerable. It is a gain of expressivity, resulting in the possibility of programming a section of the software management process that is not obviously in need of programmability, and in the necessity of scripting languages, interpreters, debuggers, etcetera for it.

    In short, it's not directly obvious to me how this isn't an ethereumization of V.

    This is defensible but it puts a great deal of pressure on the human-readable caveat component, which is also how meatworld legal contracts ended up weighing what they weigh, 100s of KB...

    No avenue of such caveat is provided.

    The gods are angry, and the gods demand blood.

    More importantly : the coder-centric perspective has got to be abandoned. This isn't a smokestack-centric ecological reform, whereby we find new things to do with the smoke, perhaps barbotage through sea so the specs of dust don't ruin the Commissoner-General's white shirts anymore. This is a de-smokestacking.

    Is it possible to zap the talmudism, or at least reduce it to manageable proportions, in favour of a small and unambiguous command set that covers common cases ?

    I expect this is possible ; and I imagine this is to be seriously considered ; but I am entirely convinced that productive discussion as well as the aforestated possibility strictly rests on fundamental cultural and conceptual shifts, of the nature discussed above.

    And in your view, "I witnessed this", "I read this", "I wrote this" are not meaningfully-distinct cases ?

    I expect they are, it's however not directly obvious ~where~ this information belongs. Your engine gear is ALSO valuable information, but it does not inform the engine crank shaft. That's just a shaft. It may inform the fuel injectors. It's complicated.

    And do you have your own, entirely different answer to the "how do I sign a formal fatwa against payload P" puzzle? I'd like to hear it.

    Not as such, and not for lack of thinking about it. But consider : patch X does not carry Y's signature. There is more fatwa than this ? How ? Why ?

    Elsewhere: Does the machine ~have~ to be a "dubiously shambling horror" ?

    Yes. I propose the #1 tenet of the new republican religion would be, that software handling complex semantics will be a shambling horror. You can't make good Unicode for the same reason you can't make good vectors here : you failed to first pin down a definive and absolute alphabet.

    And supposing that it were, then why would the dubiousness not extend to your RSAtron, your display handler, and other safety-critical components ?

    Very much BECAUSE alphabet. Display handler handles RGB ; and while this may superficially appear as extensible - it is not. Yes you may go from 2 to 8 to 16 bits per R ; but they still share the same space from 0 to 100% saturation, in narrower increments and that's all they do. Because the alphabet of display is fixed in a ~definitive~ and ~absolute~ way, jacking it becomes a discussion of what the cunt looked like before the camera, not after the DAC.

    This argument is reminiscent of 1970s airplane pilots' objection to "fly by wire" - "what if the circuit fails?" To which the inevitable counter is "what if the wings fall off."

    This is not a valid objection. I can calculate the falling off of wings, because (again - simple alphabet) ; I can not calculate the failing of circuits, because (...). For this reason the objection is spurious, and for this reason it shouldn't have been accepted. That it was accepted speaks of circumstances beyond the matter of corectness.

    If there is a long-term future for "computer", it is as a very minimal thing, that does not require "programmers" in the customary sense, only a literate operator.

    Very much so - and exactly from that perspective V aims to salvage as much as possible of life before the doom actually comes and glasses the field.

  7. > V does not exist so as to allow maximum expressivity for the coder, "here is one of the numerous internal states my head could find itself in, how do I use V to best express this". It is not crypto-powered livejournal for geeks

    The "livejournal" situation is precisely what I'd like to see a pill against, and how the thread began. I argue that mandating a "wall of text" with every serious signature leads precisely to the "myspace status" shit-flavour.

    "here's the list of acceptable internal states your head may find itself in" precisely describes the design goal of "vectorized sig" !

    > In the way this currently works, with boolean signatures, the computer makes ~no decision~ .... In short, it's not directly obvious to me how this isn't an ethereumization of V.

    Leads me to the thought : why not simply a boolean flag ? 1: "This is a signature" 0: "This is a fatwa" ? Let's say, e.g., "sign" and "negsign", respectively.

    > Your engine gear is ALSO valuable information, but it does not inform the engine crank shaft. That's just a shaft. It may inform the fuel injectors. It's complicated.

    So let's then have a brake pedal, as per above. Presently we have solely "gas".

    > But consider : patch X does not carry Y's signature. There is more fatwa than this ? How ? Why ?

    For the same reason that a well-made WOT gives you a minus sign to use, and not merely a 0, for ninjashotguns.

    > That it was accepted speaks of circumstances beyond the matter of corectness.

    Fly-by-wire was adopted.

  8. Mircea Popescu`s avatar
    8
    Mircea Popescu 
    Tuesday, 6 December 2016

    Yes, you're right in pointing out that the principal difference between wot-for-people as implemented by eg deedbot is the signed int whereas wot-for-code as implemented by v is unsigned int.

    It is not directly clear that a) one is better than the other ; b) the two cases are not actually distinct and would benefit from this distinction being maintained ; c) good quality answers to questions a and b do not depend on the current arrangement being maintained for a while longer.

  9. @Mircea Popescu

    One way to approach the problem of "ought V to have a brake pedal" would be to describe the cost (the added programmatic complexity would be minimal.)

    Can you think of a situation where having "negsig" around would lead to a hazard ? Or -- as is much the same thing -- to a logically-ambiguous statement ?

  10. Mircea Popescu`s avatar
    10
    Mircea Popescu 
    Tuesday, 6 December 2016

    Consider one simple point - as it is now, I have .seals and all is well.

    If there existed the possibility of "negative sealing", then it would behove me to have .posseals and .negseals, because obviously I can distinguish between "the stuff X signs" and "the stuff X anti-signs". Now I have more complexity in my V build tree, and therefore more possibility for error.

    What do I gain for my trouble ? If before I cared what X thought and had him in .seals, stuff he didn't sign couldn't build. If now I care what X thinks and have him in .negseals, stuff he anti-signs... can't build ; whereas if i have him in .possign stuff he doesn't sign... can't build.

    Soon enough I quit maintaining .negsigns altogether and V128 rebecomes V256.

  11. "negsign" remains the simplest, afaik, means of implementing "create a permanent/opposable record of the act of having understood and condemned $payload".

    The thing you gain for the trouble is a means for distinguishing the scenario where someone is lazy/overworked/dead and simply had not contemplated $payload, from one where he took six months to find a catastrophic bug and now goes to put the find in the perma-record.

  12. Mircea Popescu`s avatar
    12
    Mircea Popescu 
    Tuesday, 6 December 2016

    Yes. But I would expect that the correct handling of situation where one spends the 6 months and finds the bug is - he takes to the forum, and scandal ensues.

    At least that's what I see myself doing.

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.