On Wednesday 31 March 2004 08.05, Martin-Éric Racine wrote: > Would recoding names in the GECOS data from /etc/passwd into UTF-8 > solve it? No idea, if you're talking about the matching done by the Debian packaging scripts. > If your answer to this is Yes, then I think that upgrading a distro to > Sarge should also offer to upgrade /etc/passwd, /etc/shadow and > /etc/groups to UTF-8. The problem with that is that the system has no way to know what encoding is used in /etc/passwd. It could assume the locale the root user uses, or what is specified in /etc/environment, but I don't think this is safe. > > Heck, if you ask me, Sarge should be known as the "we upgrade everyone > to UTF-8" Debian release. This would imply that absolutely every > package to be released in Sarge would know about legacy encodings for > each locale and be able to recode every config file, man page, etc. > during the process of upgrading from Woody. I guess too much software is not unicode ready yet. It's a worthy goal for sarge+1, though. I work in de_CH.utf-8 - the main problems are man, which still has glitches occasionally, and AFAIK is a total hack under the hood as far as unicode support is concerned, then fonts - I had some lengthy fights with my system, and I don't want to have to set up a new system because I don't understand what's going on there. Then there is the textmode console - I don't know, perhaps the sarge installer does set up this by default, but on a potato-installed system, the text console doesn't work with unicode characters at all. The other problems are user data: I believe that the installer must not touch anything in /home - but switching to unicode causes all users to 'lose' umlauts/accents/whatever, and in case of non-latin scripts, mostly everything. So it is necessary to have very, very, very good user documentation on how to upgrade. Textfiles are easy, but how about non-text document formats? I fear there are many such formats in use without proper encoding information embedded. Already mentioned: telnet/ssh don't have a way to transport locale information over the connection - how do you tell this to users who may not be in a position to upgrade both machines at the same time and will suddenly lose the ability to use umlauts/accents/their script over a remove connection? (I don't have this problem because (i) when I write text, it's usually english and (ii) most machines I usually use ssh with are under my direct control, and so are also set up to use unicode.) Just my €.02 cheers -- vbi -- Today is Setting Orange, the 17th day of Discord in the YOLD 3170
Attachment:
pgpA8qE0SOxIz.pgp
Description: signature