On 2012.05.01 23:25, Ludovic Rousseau wrote: > Users should use a _released_ source tarball or binary installer so > major, minor, micro is enough. > Developers should use git and know what version they use.
Well, the first issue here is that pretty much all of our users are developers, so I'm not exactly sure where you make the distinction between libusbx "users" and libusbx "developers". Given that, outside of binary users, most people would recompile from source, I don't think there's much of a difference between people who recompile from a tarball and people who recompile from git. Or by "users", do you mean end-users of a libusbx based app? Also, any statement that goes "group of people X should ..." is pretty much an authoritative restriction of how people should be able to use software, which IMO goes against the aim we should have, especially with a generic library, of empowering people to use our software in any way they see fit. I very much see this freedom also applying to how they can retrieve the software, if we provide a choice. Thus, I find the idea that nobody should use git, who may want to figure out the version they use, quite ludicrous. There are plenty of scenarios where people will want to use git rather than release, one of which is the current situation, where we have a fairly high profile bug for async transfers on Windows, that is fixed in git and not yet in a release. For any such situation, it becomes paramount for users of libusbx, as well as ourselves if they come to us with a question, to be able to easily figure out the version their library is based on. > In a RERO world normal users should not use a git version. I don't think so. Less people may use a git version in a healthy RERO project, but a critical bug that is fixed in git and not yet in release very much warrant using a git version no matter how soon the release may happen. > I have no objection to provide intermediary/test versions with a mano > version. But that can be managed locally by the developer providing > the intermediary version. No need to store the nano version in git. The whole point is to enable libusbx users who may be using git (for whatever reason - it is really not up to us to FORCE them to use one mode of distribution or the other), to first help us answer the questions: "what version of libusbx are you using?". If we can't get this basic question answered with certainty, support becomes a major headache, and I am a lot less confident than you seem to be that every git user will "know" the precise version they are using. Also, what if we have a library conflict on the system, which is something very easy to produce when using a Windows DLL, where someone has both a git version and a release version, and think they are using one when they are actually using the other? The only way I see where we could afford not having a nano is if our git repository was private and nobody could clone from it, but that's not the case. Regards, /Pete ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ libusbx-devel mailing list libusbx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libusbx-devel
