On Mon, Jun 25, 2012 at 8:04 PM, Bert Huijben <b...@qqmail.nl> wrote: > > >> -----Original Message----- >> From: cmpil...@apache.org [mailto:cmpil...@apache.org] >> Sent: maandag 25 juni 2012 19:32 >> To: comm...@subversion.apache.org >> Subject: svn commit: r1353676 - in /subversion/trunk/subversion: >> include/private/svn_wc_private.h include/svn_ra.h libsvn_client/ra.c >> libsvn_ra_serf/update.c libsvn_wc/adm_ops.c >> >> Author: cmpilato >> Date: Mon Jun 25 17:32:26 2012 >> New Revision: 1353676 >> >> URL: http://svn.apache.org/viewvc?rev=1353676&view=rev >> Log: >> Teach the get_wc_content() RA callback and supporting code to handle >> MD5 checksums as content indexes, too. This *should* result in the >> benefits of this enhancement extending to users of all 1.8 clients >> irrespective of the server's pedigree, as MD5s have been in the server >> streams since ... like ... forever. > > If we want to enable this code in 1.8 we should add an index on the > md5_checksum in the pristine table with the format bump for 1.8. >
Is there already a format bump planned for 1.8? Can a format bump be avoided? If we have to bump, how hard would it be to keep everything working with 1.7 clients too (albeit with less features, slower XYZ because of whatever, ...)? Seems to me that having an additional index will not make the wc database unusable by older clients, right? I don't want to hold back on progress, and I know it's not one of our compatibility guarantees, but just ... it would make life a lot easier in big environments (with diverse svn clients (some integrated into other tools like IDE's, which all have their own release cycles)) if we could keep the wc format backwards compatible. See also this prior thread where I tried to start a discussion around this topic: http://svn.haxx.se/dev/archive-2012-04/0468.shtml -- Johan
