On Thu, Jun 10, 2010 at 11:35:35AM -0700, Shawn Walker wrote: > On 06/ 8/10 01:51 PM, [email protected] wrote: > >Folks, > >This change improves download performance by switching from HTTP > >pipelining to parallel polled connections. This gives us a 3x-8x > >performance improvement on downloads. Final testing on SWAN shows that > >this cuts 4 minutes off of a slim_install, down to 5 minutes from 9. > > > > http://cr.opensolaris.org/~johansen/webrev-16131/ > > > src/modules/client/transport/engine.py: > lines 187-189: Not your fault, but can you fix the pipelining comment > for me? It really applies to all protocols; not just http(s).
Ok. > By the way, I'm assuming that we're just disabling pipelining > completely for now with the possibility that transport may decide to > enable it based on other conditions in the future? Yes, although after bug 16203, I'm beginning to wonder if we should ever use libcurl's pipelining support again. It doesn't seem to be well tested, compared to their other code paths. It's annoying that it's suddenly, subtly, broken again. -j _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
