On Wed, Jun 26, 2013 at 05:50:58PM +0200, Stefan Fuhrmann wrote: > On Wed, Jun 26, 2013 at 2:52 PM, Stefan Sperling <[email protected]> wrote: > > > On Wed, Jun 26, 2013 at 02:46:29PM +0200, Stefan Sperling wrote: > > > I think this block of code should either enforce and fulfill the > > > API contract in the exact same way as svn_ra_get_location_segments() > > > does, or it should not exist at all. I don't think we should bother > > > with optimizations that don't behave exactly like the non-optimized case. > > > > Another idea: what about moving this special case handling into > > libsvn_ra itself, rather than having it in libsvn_client? > > > > The server-side implementation is pretty efficient already.
libsvn_ra is client-side, so I'm not sure why you mention the server. But anyway, it seems you've determined the RA API is too general to allow for this optimisation. > This has been about the extra network roundtrip (Interwebs!). > > r1496957 accomplishes that optimization now. I'll take a look. Thanks!
