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

On Mon, 2008-11-10 at 17:42 -0500, Justin Erenkrantz wrote:
> Tracking errors with a glommed request is a nightmare

In the context of Subversion, tracking errors either way is problematic.
Above the DAV layer, commits occur through the editor interface, which
has an error status return for every request. If the app assumes it's
going to get an error status in response to an editor operation, then
it's going to be sadly disappointed if the RA layer is pipelining the
requests; that information just isn't available without waiting on a
network round trip. This is already the case for ra_svn, and will be
the case for DAV whether a single request is used or a pipelined series
of PUTs.

(This is probably one of the reasons Greg Stein is no longer in love
with the editor interface.)

> - plus, knowing
> how we would almost certainly implement glommed requests, the content
> likely would be base64 XML-encoded. Ugh. -- justin

An odd conclusion, given how we glommed compound data in ra_svn. (We
did not use base64.)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-11-11 01:27:36 CET