There’s no such thing as the perfect version control tool and there never will be. That’s why extensibility is #4 on my list of evaluation criteria for VCS tools. There will always be pressure on these tools to enable new ways of working, provide more information for decision making and provide smarter integration with other tools. Over and above that, extensibility is really important for both technical and social reasons.

Technically, trying to ship a tool which is all things to all people creates all sorts of problems particularly w.r.t. reliability, my #1 evaluation criteria. Bloatware takes longer to ship each time, quality typically drops regardless, and the extra features don’t necessarily hit the mark anyway. As Mozilla has shown with Firefox and Thunderbird in recent years, it is far better to produce a rock solid core product that supports plug-ins in a documented way. Done right, the result is higher quality, better performance, and extensions that better meet the needs of the user base anyway.

Socially, a plug-in architecture increases the engagement of the community that grows around successful products. Whether open source or commercial, it takes a community to raise great software and plug-ins let that community scratch their itches in a sanctioned way.

There are different ways of tackling the extensibility challenge but the best way in my experience is by explicitly supporting plug-ins as Bazaar and Mercurial do. Other tools like Git have gone down the well worn toolkit path, and while that’s much better than having a monolithic application, I believe the plug-in path is a wiser one for a host of reasons.

(more…)