On Thu, Sep 19, 2002 at 10:17:37AM +0300, [EMAIL PROTECTED] wrote: > You seem to be ironic on that, but in fact I think it's a serious issue. > The amount of work and time wasted in parallel projects doing the same > things ten times is tremendous, and it's a shame that so much of people's > time and work is spent on things that were already done ten times - even > if in a slightly different way (by the way, one of the issues the Open > Source was called to achieve is exactly this - reduce the amount of work > duplication and raise the code reuse via source code sharing).
There's nothing wrong with reinventing a wheel, provided you do a better job than the original wheel designer / implementor. I'm not talking about "my mp3 jukebox", 17,000 of which, all exactly alike, you can find in freshmeat. I'm talking of doing something substantial, like a journaling file system or an alternative syscall tracking mechanism (ahm). Competitions breeds progress. > But seems that it's how the evolution works, and until the > projects live in purely Darwinian evolutional space without any > guiding, organising and cooperation-driving force, it's going to be > this way. RedHat is trying to be such a force - I don't know if > there would be any good out of it, but it's interesting to see they > are trying. I don't see them quite in that role. It seems to me that what they do is "pick and choose" the best of the breed. Only in the cases where they can't do that (gnome and KDE), they try to unify. It seems that preventing duplication of work is not their motive, but rather a convenient side effect. > OZ>> And on the other hand, in a world which accepts four separate > OZ>> ways of building an Unix-like system, isn't it too much to > OZ>> expect it to accept two ways of building desktop modules? > > Which brings us eight ways to build a Unix system. Don't you see > combinatorical explosion coming? Take a look at the linux distributions page of LWN sometime. You're in a twisty maze of linux distributions, all exactly alike. > It may seem (or even be) good for you - > so many ways to do things - but for RedHat and other developers it means > supporting (meaning, developing for, testing on, documenting, etc.) all > these things - with very few common APIs to base on. Actually, redhat has to be compatible with redhat. It's the independent developers that potentially suffer. Here's where the power of open source comes into play - you want it to work on your favorite architecture/distribution? Send in a patch. As for the common APIs, if you stick to portable practices (Ansi C) and standards (POSIX, LSB), with a little help from the auto tools, you too can be portable. > What RH is trying to do is to establish as large common base as > possible - and I can fully understand that. -- Muli Ben-Yehuda syscalltrack hacker-at-large
msg21909/pgp00000.pgp
Description: PGP signature
