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

Re: in NEW: utf8-migration-tool -- Debian UTF-8 migration wizard

su, 2006-12-31 kello 20:42 +0500, Alexander E. Patrakov kirjoitti:
> Martin-Éric Racine wrote:
> > su, 2006-12-31 kello 18:55 +0500, Alexander E. Patrakov kirjoitti:
> >> Martin-Éric Racine wrote:
> >>> Having merged Vincent's patch, I uploaded utf8-migration-tool to NEW.
> >>>
> >>> Since Etch will be Debian's first UTF-8 release - implying a migration
> >>> from legacy encodings for those upgrading from Sarge, which is precisely
> >>> what this tool tackles - it would be nice to approve it for Etch.
> >> Hello,
> >>
> >> 1) patrakov@home:~$ utf8migrationtool
> >> Unexpected error: exceptions.IOError
> >> Traceback (most recent call last):
> >>    File "/usr/bin/utf8migrationtool", line 40, in ?
> >>      dmrc = getconfig()
> >>    File "/usr/bin/utf8migrationtool", line 34, in getconfig
> >>      config.readfp(open(os.path.expanduser('~/.dmrc')))
> >> IOError: [Errno 2] No such file or directory: '/home/patrakov/.dmrc'
> > 
> > Works fine here, so no comment.
> This is because you have the .dmrc file. I don't (I created an empty file to 
> get past this error when writing my first mail). This file presumably 
> belongs to gdm, but I don't have gdm (I use "startx"), and your package 
> installs fine without gdm. Missing dependency?

I don't see why it would depend upon GDM, though, since Ubuntu (which is
where the package originates from) also supports KDE and XFCE.

> >> 2) The tool must handle the already-migrated case better (e.g., by adding a 
> >> line about that onto the second screen).
> > 
> > It does. Here, it says that the locale is already migrated. It also says
> > that it cannot find any files utilizing a legacy encoding.
> Yes, it does, in the case when the old locale is from .dmrc.

It's starting to look that way. :(

> >> 3) The legacy locale for Russia is ru_RU.KOI8-R, not ru_RU, and the 
> >> migration tool must handle this special case.
> > 
> > Russian is a messy case. Too many encodings, more than half of which are
> > OS-specific or otherwise standards that never gained momentum.  This is
> > further complicated by usage cases: while Unices tend to go for KOI8-R,
> > users that need to interact with Windows use CP1251 instead. Still, it's
> > up to Russian developers to add support for this; upstream simply cannot
> > anticipate every possible exception.
> OK, I temporarily take this back (because the old report was based on empty 
> .dmrc - but anyway, you could take the .KOI8-R part from $LANG). However, I 
> replace my old report with this: when the old .dmrc contains
> [Desktop]
> Language=ru_RU.KOI8-R
> the migration tool migrates this to ru_RU.KOI8-R.UTF-8 which is wrong. Also 
> it migrates de_DE@euro to de_DE@euro.UTF-8.
> The locale names generally have the form:
> ll_CC.CODESET@modifiers (where .CODESET and @MODIFIERS may or may not be 
> present). The old codeset and the @euro modifier (but probably not other 
> modifiers) must be stripped out.


> >> 4) migration of encodings is only a part of the game. The most important 
> >> part is to deal with packages that do not work correctly in UTF-8 locales 
> >> and cannot be fixed (e.g., a2ps). Since this part cannot be automated (as 
> >> nobody has created such blacklist), I suggest mentioning this obstacle in 
> >> the manual page and on the welcome screen.
> > 
> > Remaining UCS issues really belong in Etch's release notes, since it is
> > Debian's first release claiming UTF-8 support.
> Yes, they do. However, not everyone reads the release notes, 

Which then becomes the user's own problem.  

> >> Thus, I cannot recommend migration of this package to Etch in its current shape.
> And I still say this.

Sadly, I'm starting to agree. :(

However, this also means that users will need to manually upgrade their
system, which his far from ideal.  

If anyone would care to contribute code towards fixing the above, it
would still be highly relevant for Etch. At any rate, this software's
usefulness post-Etch would essentially be zero, so it's pretty much a
now-or-never case.

Martin-Éric Racine

Reply to: