Re: broken tables-big.in pango/modules/basic/

[Owen Taylor <otaylor redhat com>]

suzukis@flavor1.ipc.hiroshima-u.ac.jp writes:

> >> Could you give us the list of mapping files which you used
> >> to generate tables-big.i etc?
> >
> > Mappings/ISO8859/8859-(\d+).TXT => iso-8859-\1
> > Mappings/EASTASIA/GB/GB2312.TXT => gb-2312
> > Mappings/EASTASIA/JIS/JIS0208.TXT => jis-0208
> > Mappings/EASTASIA/KSC/KSC5601.TXT => ksc-5601
> > Mappings/VENDORS/MISC/KOI8-R.TXT => koi8-r
> >
> >The TIS mapping I used I basically created myself; I suppose I could
> >check that one in.
> 
> Much Thanks, I could not find TIS mapping table on
> the web site of Unicode consortium - I see why.
> 
> > b) the formats of the tables in the source are hard to
> >    process. (they are different for different character sets.)
> 
> Hmm, their formats are so hard to process? Just I've written a GNU
> libc charmap (glibc-x.y.z/localedata/charmaps/) version of "make-table.sh".
> (TIS mapping can be generated from GNU libc charmaps.)
> 
> However, I must apologize - the GNU libc charmaps can not replace
> all of maps provided by Unicode consorthium. The charmaps for CJK
> languages accompanied in glibc-2.0.x and glibc-2.1.x (x =< 3) are
> "reduced" style, not fully-informed. Although current glibc-2.1.9x
> informs fully, but still some maps are lacked. I should give it up,
> and should consider another method.

Ah, I see I was looking at the wrong place. (I was looking in the
processed forms of the character set maps in iconvdata/ rather than at
the character set maps in localedata/charmaps.)

While I don't have any real objection to depending on glibc-2.1.92 for
people who are rebuilding the character set maps, I still would prefer
to stick to the Unicode consortium charmaps; they are, in my opinion,
more standard, and easier for people to find.

Plus, they still are a little easier to process - compare how JIS X
0212 is represented by the two sets of data. For the glibc tables, the
perl script would have to be modified handle the three-byte EUC forms
/x8f.... while using the Unicode data, it is just a question of
sticking another file into the maps/ directory.

It probably won't matter too much either way. Hopefully, we
shouldn't have to add too many more character sets to the map tables
at this point.

Regards,
                                        Owen