Re: New encodings in makeinfo

[Patrice Dumas]

Hi,

If there is more than one really used way to generate filenames/anchors,
the html cross manual reference won't work. So I think that the default
should be what te users are most likely to use and that allows for 
cross manual references.

> _00fa_00c1_00c7_00c1_00cc_00d8_00ce_00a6-_00da_00c1_00d5_00d7_00c1_00d6_00c5_00ce_00ce_00d1.html
> 
> Zagalxni-zauvazhennya.html

That seems a lot better. I think this could be the default for filenames
for cross references. The disadvantage with regard with the current scheme
is that it allows for more ambiguity in the file names, so more unrelated
nodes could end up in the same file. All in all I think that the advantages
outperform that issue. I also think that the anchors should remain with _xxxx
to be sure to have unique anchors, having unreadable anchor names is not
as important, so the user would have to locate the anchor and cut and paste.
I also think that the current scheme should be left as is for non letter
characters, like "#'{.

If this is agreed, there is an additional step to perform, that is specify
the transliteration scheme used for all the unicode language characters,
even thoses makeinfo cannot handle, in ascii 7 bit letter+digit combination. 
Maybe the best approach would be to agree that the glibc iconv 
transliteration scheme would be used? Looking at what is in glibc, there 
seems to be some missing transliterations. I also had a look at libiconv,
and it seems that libiconv and glibc doesn't share their transliteration 
tables? 

It also seems to me that iconv/libiconv are missing a transliteration
scheme from non roman/asci-7bit characters to roman/asci-7bit characters,
although that's what we should specify...

--
Pat