Re: [Gnu-arch-users] Re: SourceForge-like repositories in an arch-centric world

[Tom Lord]

    > From: "Stephen J. Turnbull" <address@hidden>

    >     Tom> I think it's interesting to ask whether there isn't a much
    >     Tom> simpler, more robust, more flexible, more extensible
    >     Tom> infrastructure possible -- perhaps putting some of those
    >     Tom> pieces together.

    >     Tom> You know -- we can just pretend the 1990s never happened.

    > As a development infrastructure, probably.  But one thing that's
    > potentially useful about the megasites is social/legal:  they provide
    > an index of and access to full documentation for "prior art."

[Have somewhat lost the old thread but just taking the context you've
included.]

And much more, besides.  Gosh, I wish I could get GIMP running on my
box without taking a week off -- I'd draw you a picture.

Yeah, I think the ideal is a very distributed development
infrastructure and 2 orders of magnitude or greater more distros than
we currently have --- but yes, little hubs that "collect everything"
and index them ten different ways from tuesday fit into that picture
nicely.


    > Note that Freshmeat doesn't really perform the same function,
    > especially from the point of view of the PHBs who run the PTO.
    > They're looking for centralized "publishing houses" with more or less
    > verifiable histories rather than ways to access private archives.

Right.  FM is just the "departmental bulletin board, at scale".  (And,
as such, I love it.)

Beyond verifiable histories -- how's about things like certifications?
I'd like to see vendor certs and standards certs and security-bulletin
certs reduced to _extensional_ terms rather than _intentional_ (i.e.,
Oracle supports it's stuff on any platform that passes such and such
tests rather than on RH X.Y and Suse A.B).   And, "I'd like to see",
those tests reduced to a cheap service whose interface is something
like, gosh, a platform named by a set of coordinates in a global
revision namespace.

So, for example, if I'm the company making the distro for GE in
Billerica I can plunk down $500 and email the arch coords to a testing
service -- they instantiate that distro from trusted sources
(verifying with signatures :-) and run the suite -- and I get my cert
or don't.   It can be a condition of my contract with GE that I don't
install stuff without certain certs.

We're N steps away from that where N is intimidating but finite.


    > Caveat: I don't know if this will actually work for patent-busting,
    > but I can say that I've received upwards of a dozen mails from folks
    > like Greg Aharonian (sp?) requesting confirmation that XEmacs "does X"
    > and a pointer to archives where implementations of "X" can be found.

No doubt.  Eventually, software patents will suffer their own Scopes
monkey trial wherin it is demonstrated that there's a generative
grammar that takes 100 pages to express yet generates most of TAoCP
(Knuth).   Similarly for GUI and web patents.   The whole concept of
an original idea in these areas is crap -- it's all just mindless
elaboration on a few simple principles.

    > Arch's ability to sign archives is potentially very useful in enhancing
    > credibility here, too.

Yuppers.

-t