On Thu, Dec 18, 2008 at 8:42 PM, <lu...@proxima.alt.za> wrote: >sqweek wrote: >> What risk? > > Untested and/or incomplete kernel changes?
I'm not seeing the issue? We're not talking about dumping random stuff into the /sys/ of unsuspecting users here, the matter at hand is simply the availability of the code to interested parties (who might want to test and/or complete the changes). You could argue that this is the case, and interested parties need only contact the labs to get a copy of the code, and I do kind of like the approach. It has a nice personal feel to it. But it has some limitations, too, namely that you need to know the code exists to request it. You were looking for elaboration on this earlier - imagine passers by, for one: "Alright, my x86-64 board arrived! I wanted to try out some other OSes, what have we here... hmm Plan 9, seems interesting... aw, no native port! Guess I'll try losethos." It also requires a prior interest in the code - there's no chance for someone to stumble upon it and become interested that way. To be fair, you don't need to know the code exists. You could always send a mail out when starting to work on a project to see if anyone has already started... hm, which would have the advantage of keeping everyone in the loop with what people are up to, and now that I'm thinking about it would work pretty well, at the expense of some noise. Also it requires everyone's participation to work well, and doesn't deal so well with people disappearing (like the guy who did the inferno NetBSD/386 port - fortunately he had publicised his patches, so it was possible for someone else to pick them up, test them out and get them in the distribution). I'd like to illustrate my point further but this is probably already long enough, and I should be cleaning my apartment so as to avoid getting kicked out following my inspection tomorrow. ;) To sum up, I'm not trying to say that withholding code is the devil's practice and you're all going to hell if you do it and as soon as all code is public developers will fall out of the sky to write drivers for plan 9... My point is simply that there's a lot of caveats and potential obstacles that disappear when code is freely available, and I'm yet to see anyone demonstrate a disadvantage of doing so. -sqweek
