At 20:07 Uhr -0400 18.05.2002, Mark Rahner wrote: >Hi Max, I think you misunderstood some of what I was saying. >I've added comments inline below. -Mark > >Max Horn wrote: >>I understand your disappointment with the problems in the evolution >>package, [...] >>I know that it has flaws, but it definitly is useable. > >Evolution is *not* usable now because it's not buildable. It depends >on ftp://ftp.ximian.com/pub/source/evolution/evolution-1.0.3.tar.gz >which doesn't exist anymore. The current version is 1.0.5. This is >an example of the Fink-external *source* dependencies I was talking >about. I'll let you know how evolution is after I can build it. ;-)
If you had told me, I could have done something about this. But in any case, mirroring every source is not an option currently. And in most cases isn't necessary. A pity that they throw away old sources, unlike most other package publishers. [...] >>2) The tool chain for maintaining a bindist must be updated to >>support bindist. It must be improved to allow easy use, and be made >>"fool proof" (because everybody, that includes me, makes mistakes >>and it's not so nice if that means messing up the bindist) > >You have a substantial binary distribution under stable today. I assumed >that it could be supported with the same tool set. Perhaps I'm wrong. Yes you are wrong. > >>3) The people that are doing this must be trustworthy. That means, >>I must trust them to be careful (not messing up things in the >>bindist), security aware (they might not put trojans in their >>.debs, but how do I know there machines are not wide open to >>attacks because they enable root and telnet and their password is >>"foobar" ?). > >Not necessarily. I'd prefer the "people" doing this to be automated. >The binary distribution should be automatically derived from the >source distribution. > >Today (if I'm not mistaken) Fink builds require interactive answers >to questions to select among alternatives. One up-front interactive >program to manage all of the alternatives (if one doesn't exist already >- please tell me if it does) would allow the builds to be batchable. Well Fink asks you these interactive questsions at the beginning of the build. Don't see a fundamental difference there. [...] >Automated builds make a lot of sense, but where? On which machines? >Apple, are you listening? This would be a fine opportunity for >someone with a few spare Macs to make a huge difference! Get serious dude. Apple so far has shown 0 interest (this doesn't count a few single indviduals that work at Apple. They do not represent the company however). Heck, currently it seems I have no chance to support Jaguar/10.2 in advance since I don't get a seed copy of it, unless I decided to break law and rob one. So how likely is it that they would even donate/host a single machine? Maybe I see black, and I will be happy to be proven wrong, but right now I just don't see a single hint for this. >>4) We need a QA team that checks if this unstable bindist is >>actually usable, by trying to apt-get all the packages regulary, in >>various combinations (ideally, a dedicated machine with a test app >>that automates this). Otherwise we will just get as many complaints > >Yes, that QA team would be most of your users. Folks like me. There's >no getting around the fact that more users will find and report more >flaws more quickly. That's a good thing. Really! Some of these >users won't be polite, but it's easy to learn to ignore rude folks >who don't know what unstable means. The important result will be >that more packages will become stable sooner. There are second >order effects as well. More users will lead to more interest which >will lead to more developers. This ain't work. We need dedicated testers. Many things go uncaught for months, because the avarage user never notices it. E.g. think of the passwd package problem, I only noticed the old passwd package was still in stable when we were doing the 0.4.0 release. If we had a proper QA team this would have been detected far earlier. [...] > >> > You appear to be heading in that direction but with >>relatively >minor changes, you could be headed there faster. >> >>So please clearly name those "relatively minor changes" ! What you >>said so far was rather vague in terms of concrete suggestions, I >>think. > >I made or strongly implied 3 specific changes: > o Include unstable packages in the Fink binary distribution. I replied to this already, but let me repeat: Give us the resources (dedicated man power, machines, storage, bandwidth) and you'll get it. As David pointed out, Debian has 700 developers, many dedicated machines, tons of supports. We don't. > o Make Fink-local copies of all source packages that Fink requires. > o Fix the dependency problem in the evolution package. Dude that's stupid, there will always be single packages with a problem, evolution is in no way special (maybe *you* think it's a very important package, but for one person you tell me that thinks the same I can tell you ten that doN#t think so). >And I made a few more in this message: > o The binary distribution should be automatically created from > the source distribution. >o Ask Apple to help out with build resources. At very little > expense to them, it would be a huge marketing win. > o If there isn't a program that allows up front management of > alternatives so that builds may be non-interactive, it would > be a good project for a dedicated programmer with time on > their hands. For all see above and my last email. Look, I know you want to be helpful, but so far you told us nothing we didn't already know. The point is, saying "somebody should do this" helps nothing. Zero. Nada. Get somebody to *do* this (i.e. write code, provide money, servers, whatever), that will be helpful. Cheers, Max -- ----------------------------------------------- Max Horn Software Developer email: <mailto:[EMAIL PROTECTED]> phone: (+49) 6151-494890 _______________________________________________________________ Hundreds of nodes, one monster rendering program. Now that's a super model! Visit http://clustering.foundries.sf.net/ _______________________________________________ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel