log☇︎
427800+ entries in 0.279s
assbot: Logged on 05-08-2015 15:04:31; shinohai: But not connecting to anyone at all isn't
ascii_field: http://log.bitcoin-assets.com/?date=05-08-2015#1225805 << this is what i refer to as 'blackholed' ☝︎
ben_vulpes: asciilifeform: my coracle's been wedged at 367850 since last night, even with the new db locks config. last 500 lines of debug.log: http://dpaste.com/1XB21DH.txt , please let me know if more would be useful
assbot: Logged on 05-08-2015 12:35:39; mircea_popescu: http://log.bitcoin-assets.com/?date=05-08-2015#1225130 << instead of imagining the arguments you two'd better quote them. THAT much of a process we do have, and it's both explicitly stated and in daily implicit visible use.
ben_vulpes is not terrifically familiar with the git gpg integration
ben_vulpes: git produces patches tho--
ben_vulpes: <Adlai> ootb, signed patches. deeding signed patches from darcs is less work than doing so from git etc << how so? i've never touched darcs, the question comes from ignorance.
punkman: I imagine they already have people in those companies, why make it official?
assbot: China to Set Up ‘Security Offices’ Inside Internet Companies - Bloomberg Business ... ( http://bit.ly/1IpuQpx )
Adlai: ootb, signed patches. deeding signed patches from darcs is less work than doing so from git etc
trinque: I mean they all can, but well?
trinque: Adlai: it does that?
trinque: the requirement that signed and deeded patches be the primary representation of a contribution great appeal though, for obvious reasons ☟︎
trinque: wrap git or mercurial or something in hooks which enforce the constraints desired
trinque: sure, and that's another thought
punkman: I don't want another VCS system. I just need a way to import patches in my existing VCS. ☟︎
trinque: implementing a VCS isn't rocket science; it's just a matter of getting the abstractions right
trinque: anyhow one thought.
trinque: and then maybe deedbot farts out additional nav controls on the html versions of these interesting deeds with a particular sexp in them
trinque: so if you say did the structured data in the "blobs" as sexps you could sign that text easily
mircea_popescu: jurov word. there must be.
trinque: I probably missed one, and they also divide into further classifications, ref for example can be a tag, branch, stash
jurov: (currently on has to clearsign the .txt, make detached armored sig for .patch and use non-braindamaged email client to put them together)
trinque: so, git does it this way: there are Blobs, which are a k/v pair of which the k is the hash of the v; blobs can be one of: Reference, Commit, Tree, Blob
trinque: one thing that would be needed on my end is a saner URL for a given deed; it needs to just be the hash
trinque: you could implement the data structures needed for VCS atop deedbot in a similar manner
trinque: the way git handles this is that "everything's a Blob" then "some Blobs are a Tree which points to n Blobs which may be Trees or Blobs"
jurov: there must be better system to put together few files + text that already has crypto or can be extended easily
jurov: yes the signed email + extra signed attachs is a kludge
mircea_popescu: bears more thinking
trinque: no reason I can't add some html nav around that a la dpaste
trinque: to date I've tried to keep it simple, hence the deeds are plain textfiles linked to by a single index
jurov: email was chosed because it allows for structured text + attachments. not sure how to do it with deedbot
mircea_popescu: jurov i wasn't proposing they be used as is.
shinohai: But not connecting to anyone at all isn't ☟︎
jurov: mircea_popescu: deedbot has currently even less friendly archives than mailman
mircea_popescu: in the sense of retarded but commonly seen
mircea_popescu: shinohai this is "normal"
shinohai: asciilifeform, mircea_popescu: this is what I get http://dpaste.com/1889469
mircea_popescu: this readily enforces the two restraings we want enforced, which is : a) wot restricted and b) signed permanently.
mircea_popescu: the biggest headache to date for jurov/users of ml seems to have been the mailing itself. i wonder how much sense simply deedbotting messages would make. ☟︎
trinque: and I can modify deedbot- however it helps; he already statically generates his site, perhaps he could interpret metadata in certain deeds and take action
trinque: mircea_popescu | asciilifeform and mod6 and mebbe we have trinque do the graph << I'm in; seems like yet another static site generator job
shinohai: I can see 8333 from ny1.hashbang.sh tho
mircea_popescu: well... the network is pretty noisy.
mircea_popescu: but the other one no ?
shinohai: Could be I just started this one, the one I am already running returns 28670932
mircea_popescu: then why is grep returning 0. one or the other neh ?
shinohai: Where it attempts to connect but disconnects (no message in first 60 sec)
shinohai: Watching tail though I see exactly what asciilifeform describes
mircea_popescu: then it's not that. prolly gotta pursue the memory issue.
asciilifeform marvelling that no one, apparently, ever noticed the glaring omission in 'patch'
mircea_popescu: shinohai please run this : cat debug.log | grep -c "no message in first 60" and report result
mircea_popescu: asciilifeform and mod6 and mebbe we have trinque do the graph
asciilifeform: shinohai: this is a persistent problem, and i have every reason to expect that it will get worse.
assbot: Successfully updated the rating for diana_coman from 2 to 4 with note: Work on Eulora client, various tidbits over the years. Freelance statistician/coder.
mircea_popescu: !rate diana_coman 4 Work on Eulora client, various tidbits over the years. Freelance statistician/coder.
shinohai: My node isn't connected to anyone again today *sigh*
asciilifeform: incidentally it is not necessary to alter the unix patch util. the checksums can be added by a proggy which eats standard patch file and the 'before' tree, and shits - this
mircea_popescu: also i'd like to keep this conversation open, for all the folks whose time is not day atm yet. ☟︎
mircea_popescu: and the documentation.
asciilifeform: now, the above is not a magic pill! somebody will still need to write the graph traverser
asciilifeform: but now both sides of the patch (author and applier) will need the custom util.
asciilifeform: and thereby can safely snip a file after stating its checksum
asciilifeform: jurov: it was not a thing before, because there was no way to guarantee that the contents are what the patch author thought they were
asciilifeform: i will add, if it isn't obvious, that patches like the one i suggested will apply on a standard unix patch util
jurov: i like it. to be able to remove files without dumping whole contents, patch should be patched, too
asciilifeform: hence the 'never, ever include a patch in message text' thing.
asciilifeform: it was an issue with message text on ml
asciilifeform: same as a tarball would be
asciilifeform: afaik this was never an issue with patches (they were attached as binaries)
jurov: by running gpg --decrypt on clearsigned text
mircea_popescu: asciilifeform how do they handle the "armored"/asciifucks issue ?
asciilifeform: in that, having a bag of patches in hand, you could magically coalesce them into a dependency graph
diana_coman: it took me half an hour only, but yeah, I did not have to re-read the code or anything
asciilifeform: imho this would correct a fundamental flaw in the whole diff thing
asciilifeform: see the dpastes
mircea_popescu: with her there holding my hand
mircea_popescu: ~I~ applied the diffs.
jurov: diana_coman: you actually tried to apply the diffs?
mircea_popescu: asciilifeform not even sure the pile was worse.
asciilifeform: because the alternative was a multi-megabyte pile of minus signs
mircea_popescu: i dunno how you folks understand code otherwise, might be that i'm rusty or inept, but anyway.
mircea_popescu: i didn't care for that.
mircea_popescu likes, perhaps irrationally, the patch format
jurov: there may be ned to publish other signed artifacts
mircea_popescu: as opposed to what ?
jurov: though, with this in-band signalling we will limit ourselves to patch format
asciilifeform: i'm about to post an example
asciilifeform: from patches like these, one could conceivably auto-build a dependency graph
asciilifeform: thinking of a modified unix-patch util that stuffs sha512 into comment preceding each file diff
asciilifeform: as i understand, any mechanized thing really depends on patches stating their antecedents
mircea_popescu: gimme the mechanics!
mircea_popescu: yet most people to date have been killed by viruses.
mircea_popescu: gimme the mechanics, lemme worry about the people.
asciilifeform: again with the meat models. in point of fact most viruses ~don't work~
asciilifeform: the only way to guarantee that the selected set of authors adds up to a single nonconflicting merge is to only build official releases (that is, tree signed by mod6 & ben_vulpes)
mircea_popescu: and i know this can be done because it's how fucking viruses work. ☟︎
mircea_popescu: and the only time i hear of any peep from the entire shebang is when one of them has gone over and i need to come in here and start with the !rate -10
assbot: Page not found on Trilema - A blog by Mircea Popescu. ... ( http://bit.ly/1DtYhd5 )