What to do about HPSS protocol backwards compatibility for faster push?

Andrew Bennetts andrew at canonical.com
Thu Oct 18 03:06:07 BST 2007


Andrew Cowie wrote:
> On Wed, 2007-10-17 at 16:13 +1000, Andrew Bennetts wrote:
> 
> > If the request is large (e.g. pushing 10MB container of revision data), then
> > this isn't so good.  Sending a 10MB request just to get back a “sorry, I don't
> > understand” response is going to be pretty slow.  What should we do about this?
> 
> Is it so synchronous? I would have thought that after a few bytes the
> server would be able to send back an "I don't understand" even though
> the client isn't done transmitting...

See my reply to John; it would require changing the request/response model in a
way that isn't easy to implement over HTTP.

Potentially we can do this as an optimisation on media that allow it,
e.g. bzr+ssh:// or bzr://, and just not worry about (bzr+)http://... but that's
a fair bit of additional complexity for a modest gain, and would introduce
significant discrepancies in how the protocol performs over the different media.
Ideally it should work well on everything, rather than leaving (say) HTTP as a
second-class citizen.

-Andrew.




More information about the bazaar mailing list