approx and ocaml-http (was: Camlrpc and Cduce updated, new Camlgz)
On Mon, Aug 01, 2005 at 11:56:57AM +0200, Sven Luther wrote:
> On Thu, Jul 28, 2005 at 07:01:09PM -0400, Eric Cooper wrote:
> > On Fri, Jul 29, 2005 at 12:40:52AM +0200, Thomas Petazzoni wrote:
> > > I've created a new Debian package of CamlGz, a binding for the zlib and
> > > bzip2 libraries. I need this package for the demexp project, but I don't
> > > know if you are interested in uploading it into Debian.
> >
> > I'd like to see this in Debian so I can use it in approx. Right now I
> > am forking gunzip/bunzip2 processes with Sys.command (and that's still
> > faster than using Camlzip).
>
> BTW, did you ever found out the reason for approx failing to provide some
> files ? It is pretty anoying when doing d-i installs through approx.
Yes, I finally found it. APT expects HTTP/1.1 compliance from the
servers it talks to, specifically the ability to keep connections open
for multiple requests.
There is a partial workaround one can do on the client side: use the line
Acquire::http::Pipeline-Depth 0;
in /etc/apt/apt.conf.
I patched Zack's OCaml HTTP to support persistent connections and it
seems to have solved the problem (i.e., without requiring the above
workaround on the client). I'm hoping he can incorporate my patches,
so I can check in the latest version of approx I've been running.
I've also made a number of fixes to support Secure APT, since it
imposes stricter consistency requirements on the archive (and
therefore the approx cache) w.r.t. signed Release files.
> Also, i wonder if it would be possible to write some kind of NAT
> stuff to do approx calls automatically when a machine on the local
> network is trying to access the debian archives outside ?
This should be possible with iptables rules. I'll look into it. Are
the names or addresses of the outside archive machines known in
advance?
--
Eric Cooper e c c @ c m u . e d u
Reply to: