To get supported encodings by iconv try iconv -l and iso-8859-8-i is not among them, iso-8859-8 and WINDOWS-1255 are.
Besides that, encoding should be consistent in the translated he files, currently there is:
<?xml version="1.0" encoding="iso-8859-8-i"?>
<?xml version="1.0" encoding="iso-8859-8"?>
<?xml version="1.0" encoding="iso-8859-1"?>
Well, iso-8859-8 and iso-8859-8-i are not the same as far as I learned. The question is if the files under the iso-8859-8 XML encoding header are in what encoding? Probably all of the files under iso-8859-1 are really iso-8859-8-i files (as Shimi said they used this encoding for translation, but probably forgot to change the XML encoding header).
AFAIK the input encoding should be 8859-8 or WINDOWS-1255. Output encoding is set to 8859-8 by the xsl-sheets. The question: why is iso-8859-8-i used in he files?
Because iso-8859-8 is deprecated as far as I know and iso-8859-8-i (also known as windows-1255) is favored instead. So the XML headers as well as the output encoding should be changed to the used encoding: iso-8859-8-i or windows-1255. As the latter is known by xsltproc, let that be.
The entiity files:
language-defs and language-snippets might be changed, as suggested by shimi
iconv -f windows-1255 -t utf-8 to make xsltproc happy;-)
Great.
At least my tests have shown, that this works fine, at least xmllint and xsltproc doesn't complain about encodings. Can't say anything about the results, because i can't read hebrew:-)
Testing:
1. I applied iconv conversion to the two entity files
2. changed encoding to WINDOWS-1255 in the curl folder of he tree and other files in question (maybe this should be 8859-8).
Results (processed with xsltproc) can be found online at:
www.holliwell.de/he/
Can someone from the he group check this?
p.s.: changing the language entitiy files has the effect, that make test reports errors about invalid sgml characters.
Brrr. Does make test_man_gen has those errors?
Goba