Junichi Uekawa wrote: > My plan was to add support to debconf to use iconv to convert to > nl_langinfo(CODESET), from the template encoding (which is legacy > ISO-8859-whatever, or EUC-whatever). I think we're all agreed on that, but the problem is that is requires a way of knowing what the template encoding is. BTW, is there a perl module that provides an interface to nl_langinfo? > Wouldn't just having "Description-ll_LL.ENCODING:" fields work fine ? > I've got a feeling that it's doable. Well it breaks backwards compatability with older versions of debconf. You would, at a minimum, not get any translations with older versions. Some very old versions would crash on that. Of course there is a similar problem anyway; if a package ships a templates file that uses UTF-8 as the encoding and has an Encoding field and it is installed with an older version of debconf, no conversion will be done and the user might see garbage. We can avoid either problem by changing what packages depend on to something that ensures a newer version of debconf is installed of course. I kind of dislike putting the encoding in the field name. If we ever want to be able to have Description-jp.utf-8 and Description-jp.eucJP in the same template it would be necessary though. I suppose I'm willing to go this route. -- see shy jo
Attachment:
pgpcAacTNXkIP.pgp
Description: PGP signature