> I personally don't have much desire nor time to play advocacy > games. If I had wanted to do that, I would have been a lawyer. > > On the other hand, I do have an issue that I would like to see > resolved fairly quickly. It seems to me that the w32 designation > is a bit broad or limited, depending on which side of the handle > you are looking at it from. I also think that ix86 is in the same > problem. We have how many OSs that run on Intel's hardware? > I see that someone want's to start a ix86/solaris port. > And, to me, there is a pretty large difference between a cygwin > port of debian packages, and a native Wintel port, both of which > could qualify for w32. I'm thinking that I may have to repackage > dpkg, (and eventually apt), and use a cyg nomenclature, instead > of w32. But, that presents problems, if someone ports cygwin to > someother architecture, say NT on Risc, or somesuch animal. > > Basically, we need more granularity in our package naming conventions, > it seems to me.
How about using an example that has already been accepted by many people: hurd-i386 Would seem to me that perhaps cygwin-i386 would be appropriate. If no os is listed, the os is assumed to be linux in the case of Debian tools. If it s not Linux, specify it, and then put a dash and then the hardware architecture. I think considering cygwin an OS as far as the packages are concerned, seems fair. This leaves room for mingw-i386, and various other win32 ports should such things ever occour. And a native one could perhaps be mswin-i386, or something similar. Does that make any sense and seem to follow current convensions of Debian for naming ports? I imagine a Debian package system for FreeBSD would become freebsd-i386. Why treat win32/x86 any differently? Len Sorensen
