Re: [PATCH] Make svn clients indicate their operation name to backend(right now only to DAV)

Kamesh Jayachandran <kamesh_at_collab.net> writes:

>> Would it be possible for the server to detect the slave error and
>> then try the master?
>
> In fact Master is the one that gives the vague error as snipped below.
>
> <snip>
>
> "The specified baseline is not the latest baseline, so it may not be
> checked out." </snip>
>
> But it would be too late for the Master to tell anything meaningful to
> Slave(if at all such mechanisms exist).

Let me see if I've got this right. There is a MKACTIVITY that goes to
the master, then a PROPFIND that goes to the slave, and finally a
CHECKOUT that goes to master. The MKACTIVITY and CHECKOUT go to the
master because they are obviously write operations, the PROPFIND goes
to the slave because it cannot be distinguished from a read-only
operation that doesn't need to access the master. The PROPFIND
returns out-of-date info that causes the CHECKOUT to fail.

You mentioned an objection from Justin related to old clients, is this
it:

http://mail-archives.apache.org/mod_mbox/subversion-dev/200912.mbox/<5c902b9e0912220853j110d893ewe412259e48b7baed@mail.gmail.com>

Is that about old clients or is it more about things like serf that
use multiple connections?

-- 
Philip