Sunday, August 15, 2010

Team Foundation Server source control versus Subversion

There is a lot of discussion going on if TFS Source Control is good or bad. In my experience it’s just a matter of getting used to it. Once you understand how TFS Source Control works, it’s a painless process.

For a client I had to summarize some differences between Subversion(their current source control system) and TFS.


Subversion TFS Source Control
Branching in Subversion has a constant cost Branching in TFS has a variable cost based on the number of items in the subtree being branched. (content does not get branched, just the file metadata).
Subversion doesn't force you to check out an item before you start editing it. TFS needs to check out an item before you can start editing it.
Subversion does not support undeleting deleted items(after a rename). TFS supports easily undeleting deleted items. This includes being able to undelete something after a parent has been renamed.
Subversion only shows an item history. It does not show non-history related branches of an item. TFS can show you full branch topologies.
Subversion doesn't have shelve support. TFS includes shelves.
Subversion has rich offline support. TFS offline support is limited.
Subversion has server side hook scripts TFS has client-side checkin policies.
Subversion doesn't have a real rename implementation. TFS supports cyclic and dependant renames. (pending, shelving, merging, undoing, and checking in)
Subversion supports symbolic links inside of the repository that don't create symbolic links on the file system, sync just downloads items from the destination instead. (See the documentation on svn:externals) TFS has no sharing functionality.
Subversion does not track the client status. TFS has the concept of a workspace. It knows what version of the file the client has.

1 comment:

Dave Van de Eynde said...

I have to say I don't agree with your claim that Subversion does not have a rename implementation. Subversion keeps track of your directory changes and also keeps the history of items, but your client must be aware of any renames that was performed beyond the control of the client. With TortoiseSVN, for example, it's better to use the TortoiseSVN shell extension to rename files than having to reconcile those filesystem changes in the client afterwards.

Also, Subversion doesn't need explicit shelving, since you can easily move a working copy with changes to a new 'branch', commit the changes into that branch and then return to the trunk or whatever you were working with at that time. In fact, this way, 'shelving' also has history.