On one side we have Subversion and on the other we have Live Mesh. With the former we have a solid, proven, and somewhat aging version control system. With the latter, we have Microsoft’s brand new (but actually not) Internet-friendly synchronization masterpiece.

With Subversion, you can easily commit changes, view different versions, and generally maintain your source code. Several different clients can all get updates from and commit changes back to the server in the typical client server model:

With Mesh, multiple computers (including an optional virtual desktop call "Live Desktop") are automatically kept in sync. Any number of clients can be added to the user’s Mesh network:

This is great and all, but each technology by itself is limiting. With vanilla Subversion, a user can work on a set of files at each client, but if he wants to continue working elsewhere on the changes that he was previously working on at his desktop, he simply can’t. With pure Mesh, a folder can be synchronized between multiple computers so that sets of working files are always and immediately available. However, in this case there’s no versioning. There’s no source control. The current version is the only version. So how could these two technologies be mashed up? There’s a couple potential ways…

The Gateway Method

Each Subversion client is an entry into the user’s Mesh network. The dual purpose machine acts a gateway between the two logical networks:

The gateway would sync the source from Subversion and that source would be synchronized across all Mesh nodes. The working set would always be available. So everyone’s happy, right? Well, not quite. The only way to commit the working files back into Subversion is through the gateway, since it’s the one computer that is both a Subversion client and a Mesh node. The other Mesh nodes have no direct access to Subversion other than through the gateway. This also means that the downloading of any other files the user might need could only be done on the gateway as well. Doesn’t sound like we got it yet.

The Full Synchronization Method

The other way is to force each and every Subversion client to be a Mesh node:

In this manner, the working set of files would always be available and the user can commit from whichever computer he’s on. An update in Subversion to grab the latest source would then be replicated across all Mesh nodes. Everyone’s happy now, right?

For starters, you have a chicken and egg problem here. If you set up Live Mesh synchronization first, then Mesh will download the files into the synchronized folder. But then when you set up the Subversion client, the first update will fail because the files and folders already exist. You can’t just delete the files, because Mesh will automatically delete them throughout your Mesh network. Not good. However, this is easily fixable by just signing out of Mesh temporarily, updating Subversion, and then signing back in to Mesh.

Making any changes on any of your Mesh computers will automatically be synched to the other computers. This makes updating Subversion obsolete. The updating is constant and automatic. You still have to (or at least you still should) check in your changes so that subversion can version the files correctly and so that others can see your check-ins, but if your source code repository only has a single user (you), you really aren’t forced to commit your changes.

And all this would be true in a perfect world where software works correctly and bug free all the time. Unfortunately, we don’t live in that world. After performing the Mesh and Subversion client installs on two computers, Mesh seems to be an in infinite loop. Mesh is continuously deleting and resynchronizing the files I deleted while offline from Mesh so that I could synchronized them through Subversion.

Switching to "work offline" didn’t stop the loop. The only way I could terminate it was to delete the problem files one by one… all of them. The unresolved conflict loop then stopped, however my files were no more. What to do now? How about updating via subversion a second time? Unfortunately, that leaves us with the same unresolved conflicts loop.

While this all seems like it should work perfectly well, it doesn’t. At the end of my testing, I have two computers using Live Mesh and working as Subversion clients. One of the computers is synchronized with my Live Mesh Desktop, but the other computer is actively  not synchronized. This seems to be a bug with Mesh since Mesh should synchronize all changes no matter how they’re made. Maybe this process would have worked with modified or rearranged steps. I don’t know.

So what have we learned here today? Well, the full synchronization method is too error prone and problematic. The gateway method would probably work fine for most people, although I proved that Mesh still has some synchronization bugs, so it may not be ready for prime time just yet. Either way, if you go about attempting to set this up, tread carefully, do your Subversion updates with care, and watch for Live Mesh synchronization issues. And don’t cry when your files mysteriously vanish. You’ve been warned…