Date: Wed, 23 Oct 2002 10:55:24 -0400 From: The Anarcat <anarcat@anarcat.ath.cx> To: Gerardo Enrique Paredes <gerardo@musiquera.com> Cc: Gerardo Enrique Paredes Mancia <g.paredes@unitec.edu>, freebsd-binup@freebsd.org, Murray Stokely <murray@freebsd.org> Subject: Re: Fwd: Breaking up base (was: current state of the binup project) Message-ID: <20021023145524.GC321@lenny.anarcat.ath.cx> In-Reply-To: <20021023105216.GA28355@musiquera.com> References: <200210230212.04522.g.paredes@unitec.edu> <20021023105216.GA28355@musiquera.com>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --]
[wrapping lines]
On Wed Oct 23, 2002 at 03:52:16AM -0700, Gerardo Enrique Paredes wrote:
> On Wed, Oct 23, 2002 at 02:12:04AM -0600, Gerardo Enrique Paredes Mancia wrote:
> >
> >
> > ---------- Forwarded Message ----------
> >
> > Subject: Breaking up base (was: current state of the binup project)
> > Date: Tue, 22 Oct 2002 13:46:18 -0400
> > From: The Anarcat <anarcat@anarcat.ath.cx>
> > To: Murray Stokely <murray@freebsd.org>
> > Cc: GERARDO ENRIQUE PAREDES MANCIA <g.paredes@unitec.edu>,
> > freebsd-binup@freebsd.org
> >
> > On Sat Oct 19, 2002 at 10:54:17AM -0700, Murray Stokely wrote:
> > [...]
> >
> > > the same mechanism. We were not interested in changing the package
> > > format, because other groups (i.e. libh / openpackages) are more
> > > focussed on that. We were focussed on updates, delivery, and breaking
> > > up the base system into discrete components.
> >
> > Really? I've been dreaming about breaking up base for a long time
> > now. Have binup started any work in this direction?
> >
> > I have a "paper" I started writing that I've been thinking to send to
> > -arch for a while now, should it be sent here instead?
> >
> Well, maybe i am wrong, but for doing a binary update everythin is
> almost laid, at least that's how i am trying to make it, this is the
> general approach i am gonna take:
>
> - the client contact the server and ask to see if a profile is
> available, let the profile be FreeBSD-4.7-RELEASE then asks for a
> certain "distribution", let' the distribtuion be doc.
> - The client now request the doc.aa, doc.ab, doc.?? chunks from the
> server, the server get them from their location on disk and send
> them to the client using sendfile(2) as Murray suggested.
>
> - Happilly the client get the chunks then checksum them and follow a
> silimar approach to install them (i.e: cat doc.?? | tar --unlink
> -xpzf - -C ${DESTDIR:-/}) as the current sysintall binary update
> procedure does.
>
> As i said before, correct me if i am wrong but this an "It Works"
> way of doing it , but i may be wrong about this is the way the
> project wants to do it. Please correct me if i am wrong and expect
> patches by the next week.
Sure it works.
The point is that this could be improved, but not necessarly on the
binup side. The client/server protocol is interesting and should
provide a powerful tool.
But everything about the doc.* chunks and the client checksumming and
installing them is not part of binup. It's part of the package system.
My point is that binup should be package-system agnostic. I would have
a directory of "packages" (whatever the format) associated with
profiles, but wouldn't impose a file format.
A.
--
Stop the bombings.
Stop the murders.
Anti-war.
[-- Attachment #2 --]
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.0 (FreeBSD)
iD8DBQE9trhbttcWHAnWiGcRAkqGAJ9wCORF1y9MQQuEkjnIGxCCHIeTygCgkMOo
qOXbr8xCTHjGZ16wVR88wUQ=
=rCDF
-----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20021023145524.GC321>