350600+ entries in 0.228s
mod6: jurov: oh ok, you were gonna have
the ML dump
the vpatch & seal out
to a second mirror?
jurov: mod6 if it can validate a signature (by name, NOT by scanning all files), it will add
to its v mirror
☟︎ mod6: its a good start
to just not mangle
the filenames --
thats all set now right?
mod6: so what are we
trying
to achieve here?
jurov: that does not mean it can't be
tidy
jurov: it will publish and otherwise ignore it, I will
tell everyone
to add it
to
their v repository by hand.
ascii_butugychag: what i say is
that REGARDLESS of
the names, if
THE CRYPTO is valid, it is a VALID patch/seal
tuple.
ascii_butugychag: in EXACTLY same way as my
trb node does
THE SAME amount of number crunching whether it hears a block from gavin's node or from mircea_popescu's.
ascii_butugychag: this means
that if jurov's box has
to verify my patch against every known sig every single
time it presses
to post
to www, SO SHOULD MINE
ascii_butugychag: this means
that at no point do ~i~ get
to do less verification ~because of something
the mirror host does~ - e.g., verify mailed in patches
ascii_butugychag: the other
thing i oughta mention is
that imho a core principle of v-ism is
that it is impermissible for
trust
to be implicitly delegated.
mod6: any way
the real sig name is currently "bitcoin-asciilifeform.1.vpatch.mod6.sig" not <+jurov> only
the question what is
the '1' in bitcoin-asciilifeform.1.mod6.vpatch.sig doing
there
ascii_butugychag: because i don't believe
that any project has any business cancerously growing patches until
this
turns into a serious boojum.
ascii_butugychag: the O(N^2) instrinsic runtime of unknown-patch-bag+unknown-sig-bag is something i realized from
the start
☟︎ jurov: so if
there is a RFC someday you're against specifying filename convention?
ascii_butugychag: (or rather, a slightly improved 'v' will run, existing one requires patch name
to remain same)
ascii_butugychag: v will run if you rename
the patch
to 'fuckyou' and
the sig
to 'fuckapig'
mod6: a long
time ago. we had a different convention.
jurov: only
the question what is
the '1' in bitcoin-asciilifeform.1.mod6.vpatch.sig doing
there
jurov: nevermind, i said, i did not
though
to be som paramount
PeterL: ok, makes sense
to me, why change it?
ascii_butugychag: the .asciilifeform.sig IMMEDIATELY
tells you
that asciilifeform signed
the
thing before
the .
jurov: instead of *.vpatch.*.sig you have *.vpatch.sig filenames which are a bit easier
to work with
jurov: also, what
the 1 left over after splitting by '.' does
there?
jurov: ok, why can't
the .vpatch and author fields swap?
ascii_butugychag: i wrote v
the way i did so
that ALL patches and ALL seals can coexist on my disk and it be HUMAN-obvious which belong
to whom and what.
jurov: but
this is Not Possible, now I have
to parse
the patchname out, and use
that
to look for sigs
jurov: i want filename.vpatch
that i can just
take and slap .sig in
the end of it
to find a signature
ascii_butugychag: if you want
the filenames
to be garbage, you will have O(N^2) evaluation.
ascii_butugychag: and now jurov has something
that does not ? so fix ~it~ plox ?
jurov: no it maps
to patchname.vpatch.<unknown>.sig
jurov: because i need
to match it
to patchname.vpatch
ascii_butugychag: i could even see
the argument
that 'signer' oughta be a gpg fp
jurov: how will sane os solve
this?
ascii_butugychag: why
tie
to filesystem oddities
that may go away when we get a sane os ?
jurov: or
try
to match each vs. each
ascii_butugychag: and realize
that we can set all
the filenames
to 'fuckyou' and v will still work
jurov: someonse sends 3 patches with 3 sigs, I either have
to get
the match with gnarly regext above
jurov: and how does ml match signarute
to vpatch?
mircea_popescu: if i want
to represent alf's signature as "asciilifeform", "ascii_butugychag" or more or just a subset, it should be up
to me.
ascii_butugychag: the correct way
to do
this is for
the first seal
to be deedbotted
to produce
the attribution of
the author.
jurov: you say noone will ever need
to?
ascii_butugychag: recall how mircea_popescu argued
that it makes no sense
to mechanically distinguish between author and signers?
jurov: then use dot after authorname,
too
jurov: is it
too late
to use some better delimiter
to extract
the parts?
ascii_butugychag: jurov:
theoretically you can avoid using
the name prior
to .sig, but
then you have
to check ALL seals agains ALL patches ALWAYS and
this is O(N^2)
☟︎ ascii_butugychag: these, ideally, will NOT always be
the same, i keep
trying
to encourage folks
to read and sign MY patches (and
that of others)
ascii_butugychag: jurov, mod6: i designed
the naming convention
this way deliberately
jurov: so imma have a regex like ([^_]+)_.*?\.vpatch\.(.*?)\.sig
to find out something
to feed
to v?
mod6: and
then v can differentiate
them easily
mod6: which, i
think is fine because
then i can sign
that patch and call it: asciilifeform-kills-integer-retardation.vpatch.mod6.sig
mod6: that's how
they need
to be.
jurov: oh my, who designed
this?
jurov: or V requires author's name in
the suffix?
jurov: one is .anything ,
the other one .anything.sig
jurov: why
the second .asciilifeform ?
mod6: that's how
they need
to be
jurov: not it is not related
to v
mod6: <+jurov> oh when i see it:
trinque and everyone, pls send detached signatures as <name>.sig << for V
to work properly
the sigfiles (seals) need
to be named
the same as
the vpatch filename with a .sig on
the end. is
that what you're saying?
jurov: ml currently passes all attachments regardless when
there is valid clearsigned
text but it may not always
the case
jurov: oh when i see it:
trinque and everyone, pls send detached signatures as <name>.sig
shinohai: I'm gonna give
those a go
today on
the server, it's idle since I got latest patched.
mod6: Attention
TRB
Testers: If you want
to help
test, please
take
the
time
to build
trb via
trinque's makefiles here;
http://therealbitcoin.org/ml/btc-dev/2016-January/000190.html (Should be basically getting & verifing
the
tar ball; plus setting up a ~/.wot dir with keys for V
to use) --
then a `make` in
the directory. Please report your findings.
Thanks.
☟︎☟︎ gribble: Chaang-Noi was last seen in #bitcoin-assets 1 year, 48 weeks, 6 days, 8 hours, 16 minutes, and 27 seconds ago: <Chaang-Noi> might have
to sue em
assbot: Logged on 05-05-2013 09:52:31; mircea_popescu: perhaps gavin, but only on
the condition
that he keeps quiet a lot on most issues.
mircea_popescu: there's a simple privacy breaking
thing where a site feeding you a bunch of favicons can figure out where you've been by looking at which you load.