[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[locale] Re: Q about XMMS i18n bug

On Wed, 29 Aug 2001, Pablo Saratxaga wrote:

> Kaixo!
> [I Cc: havardk@xmms.org]
> On Wed, Aug 29, 2001 at 07:48:40PM +0700, Dmitry Yu. Bolkhovityanov wrote:
> > 	I'm writing to you since you are mentioned in XMMS's "About" as
> > i18n programmer.
> I added the first i18n support, but haven't worked on it since then.
> > 	As you probably know, there is a funny situation in Russia -- Unix
> > uses koi8-r, while Windows and hence all .mp3s -- cp1251.  So, all the
> > ID3 tags are in cp1251 and it became a practice to set a *1251 font in
> > Preferences|Fonts|Playlist.  
> > 
> > 	This worked in xmms-1.0.1, but at least in 1.2.4 this behaves
> > badly: it seems like XMMS applies its own translation to the text to
> > display, so that titles become absolutely unreadable.  And what happens
> > doesn't depend on locale -- the result is identical under both
> > ru_RU.KOI8_R and en_US.
> Well, the problem is that CJK people needs fontsets to display the
> titles, and fontsets need to comply to locale info.

	Ooops!  Still don't understand how usage of fontsets could lead to
such strange results ;-)  (1251 uppercase remains untouched, while
lowercase are mangled).

> Have you tried with ru_RU.CP1251 ? It should display the titles right
> I think.

	Nothing changes with LANG=ru_RU.CP1251 (except that Gtk uses very
ugly (and definitely wrong) fonts ;-).

> Now, it *is* a problem; but not an xmms bug, a lack of vision of the
> mp3 format :-( (it should have used unicode!).
> Maybe the solution will be to add the possibility to set the encoding
> used by mp3 titles, and have xmms use iconv internally to convert, so
> I/O (eg typing title names and displaying them) continues to be done
> in the user locale, maybe the title from the mp3 should also
> be checked to see if it is in utf-8 (as it is very detectable encoding,
> and will probably gain acceptance in the future).
> So, Havard, do you agree to add a "charset encoding of mp3 titles" droplist?
> (maybe with an "automatic" value also, that will preset it; eg if the
> user encoding is cyrillic, then set to cp1251, if it is japanese, set
> to shiftjis, etc.)

	That's exactly the way which got consensus in russian locale
maillist.  This problem affects many apps, most known are CADs, which save
text as untagged character streams. 

> then have iconv() be used to read/write mp3 titles (when reading them, it
> should first be tried as utf-8, eg do a utf-8 -> locale encoding conversion,
> if it is ok, it was in utf-8, if it fails, then do a conversion using the
> user suplied encoding for mp3 titles).
> the charset encoding used by the user can be obtaiend with
> nl_langinfo(CODESET);
> when converting with iconv add "//TRANSLIT" to the target encoding,
> so if there are some non convertible chars they will be translitered
> (eg, when converting from cp1252 to iso-8859-1 you lose the 'oe', but it
> could be translitered as 'o' and 'e')
> PS: could you tell me an url with a (small) mp3 file with cp1251 title,
> so I cna test? thanks

	I couldn't quickly find a small file, but
http://ulpc08.gsi.de/~yakimenk/alsu-zimnij_son.mp3 is ~4M and since it is
in Germany, connectivity should be good.  Mpg123 displays the right title.

	  Dmitry Yu. Bolkhovityanov
	  The Budker Institute of Nuclear Physics