'Nested' working copies
From: Kenneth Olving <Kenneth.Olving_at_frameworx.com>
Date: 2003-09-23 20:01:11 CEST
With 0.27 I played around a bit with svn:externals. I then noticed that it failed to recurse properly. Apparently this has been fixed in 0.29 (#1428). Ok, so that's good.
However, the externals stuff does not quite fit my needs. It is great for tracking a configuration 'mix' of things, but only if the things it brings together actually are things stored in svn repos. My model and tool also allows for tracking 'derived' (prebuilt) objects such that they can be used instead of the source they came from. These objects are immutable and makes no sense to store into a VC system - thus, those things I download and cache as static content from 'somewhere'. Whatever - I can go into more details later if needed.
The Book describes that 'you could certainly setup a nested wc by hand ...' etc. In fact, that's what I'd like to do (since my tool already handles the nitty-gritty details, the 'user experience' is still good). But: will svn operations like svn status correctly recurse in that case, like it would if it had been set up by svn:externals stuff? I know, I know - I could simply test it (and will) but maybe someone has a definitive answer right off the bat...
A follow up: in the case of svn:externals which brings in >1 physical repo (or a manually nested repo as described above), will a commit be atomic across repos? I doubt it at this stage since I haven't seen any mention of TPC and distributed txns but...technically Berkeley DB seems to support XA. Is there any plans at all to provide XA capability in svn? Something like that would be absolutely fantastic, not only in the >1 repo case, but also to coordinate between, say, a repo commit and entries made into some XA aware db.
Everyone else has a cool sig, so I have a feeble one.
This is an archived mail posted to the Subversion Users mailing list.