2010-11-04 I use mercurial for version control. During the last year I heard from several people on the Internet, especially in Debian, that they switched from mercurial to git. The main reason is that it is more widely used and the community is larger, besides, it has more features. Mercurial is appearing to lose the race. I think mercurial had chosen its position badly. Wasn't right from the beginning, when git was chosen for Linux, the Kernel, clear that it will make the race? If not mercurial would have been magnitutes better, how can you compete with such a force. And that's the point: Merurial is just like git -- they cover the same niche. If there are no important reasons why someone should mercurial people will choose git, because it's the number one (thanks to the Kernel). The only chance for mercurial to not be ``eaten'' by git is plac- ing itself in a niche where git is the worse choice. Unfor- tunately, the mercurial community did right the opposite. They added all features that git had, in the belief to keep up with git this way. But this way they limited the success of mercurial to the nearly-impossible case of overtaking git. Instead they should had better strived for otherness. Of course, they would have abandonned the idea of ruling the world, but this was/is an illusion anyway. Some of you might consider it a bad example, but take the editor ed. Although it is technologically outdated, it is still avail- able and used. The reason are threefold: (1) It lives in /bin and therefore is always available. (2) It works with every ter- minal type, no matter how broken. (3) It requires minimal Bandwidth. These properties keep ed alive. I agree that the life of ed is not what someone wants for mercu- rial, but the example is only to show the general concept. You will probably be able to find a better example your own. What affects me, by the way its evolution went, masqmail lost some conceptual attraction for me. It's not neat, clean, simple anymore but grew large. Now it's just like git -- full-featured. It offers more features than neccesary, thus making its own use a complex task and hence drawing attention from the actual task: programming. It also has some module system that makes different versions incompatible. I don't want to talk against such things, but git already offers all that, why does mercurial need it too? In concequence, and to my regret, the simple mercurial that I used to know from pre-1.0 times is gone. But that exactly had been the reason why I had chosen mercurial back then. Today I worked on nmh, by continuing my ``code walkthrough''. I'm still walking round the intricate MIME handling stuff, which is the part where I want to work. The reasons is that I feel the need for more knowledge of nmh -- how everything interacts -- be- fore I fiddle in the intricated parts. Besides, I improved man pages in minor ways and modified (im- proved IMO) mhparam(1). I'll send some patches to the mailing list soon. I also worked a bit on masqmail and pushed my recent changes on- line. The next release isn't far as it seems. :-) But I need to dig further into the -oem issue and solve it first. http://marmaro.de/lue/ markus schnalke