Re: HTTP protocol v2: single commit request, or many PUTs?

On Mon, 2008-11-10 at 09:28 -0600, Ben Collins-Sussman wrote:
> * opens the door for pipelined PUTs, the way ra_svn does it
> * many fewer network turnarounds

These are the same thing. :)

"Pipelining" just means transmitting a bunch of requests all at once
without waiting. From a network usage perspective, that's very close to
the same as glomming them together into a single request.

So, you're not going to see any performance benefit from pipelined PUT
over single request. Parallel PUTs might be faster in some cases (see
my longer post about this), but any speed gain is not going to be worth
the architectural complexity, and it would frequently come at the
expense of other users of the network.

I think it comes down to code complexity. Greg Stein might be right
that the separate requests are easier to understand and code for, or he
might be wrong. You're the one working on the code at the moment.

(My very basic inclination is that single-request is better because it
lets the code use the libsvn_repos commit editor. But that could be
completely off-base.)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-11-10 21:47:29 CET