I’ve been predicting that more and more products will begin to include built-in version-control features, and I am seeing that prediction come true in a variety of products. so, I’ve been wondering how all these different tools might further interact in meaningful ways. But, I find that tools often evolve as the associated human work flows acclimate to the environments created by the tools. That is, an idea like the one that I’ve suggested won’t begin to gel until many others have the kind of work flows that call for such activities. Does that sound vague? Well, this idea is still pretty vague, and I guess that’s my point. Thanks for the response.

If it allows you to work offline, and commit against a central repository, well, Bazaar for instance can do this against SVN and other stuff. The only difference here is whether you store the whole history locally, as there is one use case in which you don’t want to do this. Again, Bazaar gives you “checkouts” without history, so it can act right now as a SVN super-client (i.e. local commits, centralized development.) That’s basically what I would expect from a hybrid client — but I wonder how you think such a client would look like?

The one use case in which you don’t want history is if you have large binary blobs in your repository (game assets, for instance.) In this case, you really want a centralized system or at least store all history on the server side instead of potentially duplicating it among all your clients; and you also want to have sparse checkouts, something which is much easier with centralized clients as you don’t have to figure out which history is relevant for this subpart of your repository. This is one (and maybe the only) area where I don’t see a real distributed alternative showing up. For all other stuff, I guess most people will migrate over to DVCS over time, as it gives you lots of benefits at basically zero cost.

Small note: I also used git and mercurial, and especially with git, there is some cost in moving from a centralized system over as it has very different concepts. Bazaar on the other hand makes it very easy to switch, especially as you can configure it like SVN for the beginning (use bound branches against a central repository only) and slowly move away from this (locally branch/merge for feature development, start merging between clients, finally move branches onto the server.) The only downside here is that Bazaar does not yet provide an easy way to delete branches on the server from Bazaar itself (you have to switch to a SSH client and do it there …)

http://ubuntuforums.org/showthread.php?t=26623

Another Voice :

Hey, nice little article, what about a little working example? thnx

Is there anything in particular which you can’t get working? Setting up an example project is something I don’t like too much, as people tend to copy & paste without understanding what they do …

Never tried STLport, I often hear that it’s supposed to be faster than the Dinkumware STL, but I’d really like to hear of some “success stories”

Initially I wrote a simple hashmap with c++ standard library, and my P3500 running debian (g++4.3, libstdc++) would outperform my quad core 4gb win xp machine (vc7.1 + standard library) :-/ whooops.

stlport on win32 is also generally faster.