2008/2/7, Tim Ellison <[EMAIL PROTECTED]>: > Alexey Varlamov wrote: > >> I have a natural aversion to svn:externals since it messes up the SVN > >> version control. When you copy a resource with an externals property to > >> a new branch, the branch has a copy of the property (not what it points > >> to). This makes it too easy to modify the HEAD when you think you are > >> working on a branch, and makes the branch mutable from the HEAD. > > > > The SVN manual covers this issue; they say "commit" command > > specifically ignores external subdirectories unless they are > > explicitly pointed. > > Right, but unless you know you are in an externals subdirectory it is > too easy to, say, cd into the java 6 branch class lib concurrent module > standard directory and make a change that gets picked up by the java 5 > branch. We know about that case, but if there are a lot of externals > pointing around the place you can miss these hidden dependencies.
Using HTTP protocol should be good enough preventive measure for us. BTW I recall Chris complained once that he had troubles checking out HTTPS-based externals in a HTTP-based workspace, I think we should fix this. > > > BTW similar issue may happen in a "mixed" workspace with > > subdirectories switched to different branches. > > Yes, and for the same reason I would also be unhappy if we switched at > levels below the working_* directories. It just gets too hard to manage > multiple branches of the code. > > Regards, > Tim > >
