On Fri, Jan 18, 2008 at 08:19:05PM +0300, Dmitry Kurochkin wrote:
> 2008/1/17, David Roundy <[EMAIL PROTECTED]>:
> > On Thu, Jan 17, 2008 at 04:07:26PM +0300, Dmitry Kurochkin wrote:
> > > 2008/1/17, David Roundy <[EMAIL PROTECTED]>:
> > > > What changes did you have to make for curl pipelining? These are 
> > > > changes in
> > > > libcurl itself?
> > >
> > > Yes, these are changes for "better" (current pipelining option is
> > > pretty useless) pipelining in libcurl. Here are 2 threads on this
> > > topic:
> > >
> > > http://thread.gmane.org/gmane.comp.web.curl.library/17197
> > > http://thread.gmane.org/gmane.comp.web.curl.library/17378
> >
> > Oh yay, so you're now an expert on how pipelining works in libcurl!  :) Any
> > chance we can get something like priority queuing? (Now that you know the
> > code better...)
> 
> I am not sure priority is needed in libcurl. But I plan to implement
> it at darcs level as you proposed earlier.
> 
> Idea is to have a pipe length limit in darcs, we download no more than
> this limit at a time. Darcs stores a list of urls which were requested
> for download, but not yet passed to libwww/curl.
> When new url is requested for download we just append it to this list
> to beginning or to end depending on priority.

Yeah, that sounds good.  I was just thinking that if it was easy (and wise)
to implement in libcurl, that might be a good option.
-- 
David Roundy
Department of Physics
Oregon State University
_______________________________________________
darcs-devel mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-devel

Reply via email to