Excerpts from [EMAIL PROTECTED]'s message of Tue Sep 09 13:03:36 +0200 2008:
> On Tue, Sep 09, 2008 at 14:58:31 +0400, Dmitry Kurochkin wrote:
> > - older than 7.18.0 - no pipelining
> > - from 7.18.0 to 7.19.1 - pipelining support is enabled, but not used by 
> > default
> > - from 7.19.1 - pipelining support is enabled and used by default
> 
> Thanks for explaining that!
> 
> > I hope the idea of using tags for making archives will be implemented
> > - that would give very good speed up.
> 
> Yes.  I think Nicolas might be interested in working on this.  In any
> case it's something I think we should do during the sprint at the
> latest.  Making darcs download fewer, larger files should be a big help
> where latency is the issue.

I meant that it's a good task for the sprint, I'm also willing to contribute
on it.

While we're at it, I've some questions about the design of it...

* Will this mechanism be an alternative to have all individual patches?

* Do we want it to be usable on http?
    Seems a dumb question, where we want to say "yes of course".
    But if we answer yes, to the first one it's less obvious,
    it would be unacceptable to download a big file because one only need
    a small part of it.

Solutions?

1) One keep individual patches and that's up to the client to download
   smartly

2) One allow packs to incrementally replace individual patches and
   provide a custom protocol to ask a given set of patches that will
   allow the server to look at it's packs and returns what's needed.

All the best,

-- 
Nicolas Pouillard aka Ertai

Attachment: signature.asc
Description: PGP signature

_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to