Hi, On Thu, Nov 13, 2008 at 10:02:32AM +0100, Gerfried Fuchs wrote: > * [EMAIL PROTECTED] <[EMAIL PROTECTED]> [2008-11-13 > 01:55:40 CET]:
> > (There are many other problems though: Aside from the broken > > entities, screen positions are miscalculated, resulting in misplaced > > link highlights and stray characters at line ends. Also, if the > > input charset differs from the terminal charset, things won't work > > at all. All this requires proper charset support to fix, which is on > > the top of my ToDo list. However, I'm still not sure how to > > implement this, so I doubt I could do it in time for lenny, even if > > the release managers would actually accept such a late change...) > > I'm not sure, but shouldn't libiconv be able to help you here? Yes, libiconv is clearly the right tool for the actual charset conversion. There are a lot of open questions though. At which point should the conversion be done? How to determine the right document charset, and turn it into something iconv understands? What do we need to adapt for the fact that we are dealing with different charsets? How to properly do line wrapping in view of multibyte characters? (And wide characters too, if we want to do it really properly...) How does this interact with characters not coming directly from the document, but rather generated internally, like entity references, or the various helper characters inserted in the output? The overall amount of code required is probably not big; there are just a lot of things to consider. (Or else I would have implemented it a long time ago :-) ) Perhaps I should split this in two tasks: first implement only proper handling of utf8 documents in utf8 locales, leaving actualy charset conversion aside for now... Anyways, I guess this kind of discussion would be more appropriate on the netrik mailing list :-) BTW, I integrated your patch into upstream CVS -- will probably release it as 1.16.1 one of these days... Many thanks for looking into this :-) -antrik- -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]