On Fri, 2003-01-17 at 17:49, Manoj Srivastava wrote: > Hi, > > Sorry for the late entry into the discussion. I am > comfortable with making the changelog UTF-8 only, but file names in > pure UTF-8 perhaps is premature. (मनोज्.conf, anyone?).
Please see my second proposal (the third in #99933), which drops the recommendation for programs to create and read filenames in UTF-8. Of course, this doens't make the problem go away; we will still have some programs creating filenames in UTF-8, and others in the locale charset. > Indeed, > until we have a wider deployment of a font that has a decent > coverage of UTF-8 glyphs (haw many of y'all can read ሰማይ አይታረስ ንጉሥ > አይከሰስ። ?), I admittedly can't; Evolution will have somewhat poor support for non-Latin Unicode until it's ported to GNOME 2. But note that UTF-8 will work quite well I think for users of Latin and East Asian languages, because we do have good, widely available free fonts for those. > perhaps we should stick to pure ascii file names, if we > must have policy take a stance about file names at all? First of all, I strongly believe policy should have a stance about file names. People will want to have packages including filenames with include non-ASCII characters. There are something like 15-20 in Debian now, and that number is probably small because of this encoding mess. And if those packages want to, we need a defined encoding for doing so. I think it is pretty obvious that UTF-8 is the only sane choice. Second, people will want to create files with non-ASCII names on their own computers; it would be bad policy specifed one charset, but users were creating files in another. But we can leave this issue aside for now. > That is not saying anything about programs that deal with > file names having widechar and encoding support, etc. I feel, as > integrators, we must follow, rather than lead, the majority of the > producers of the software components we integrate. I understand your position. In my latest proposal, policy is silent on the encoding for file names to be used by programs in general. We can fill that in later (and I think we will be filling it in with UTF-8), but I'd really like to set up the Unicode infrastructure in policy now. This will also have the effect of letting people know our intentions now, and hopefully spark a few upstream authors into adding Unicode support.
