On Tue, 28 Feb 2006, Hans Hagen wrote:

Werner LEMBERG wrote:

...

...

...

I thought I'd pass on this recent note from Werner about pdftex generating a much larger file than dvipdfmx. [...]

thanks a lot. Very interesting! This was done with 1.21a, but even the new 1.40.0-beta is not better.

This is the very reason why pdftex is a bad choice for documents written in a CJK script.

indeed, although i remember that thanh looked into it (or was going to look into it)

it should be put high on the agenda

it is there, now :-) The /Encoding in the embedded Type1 font file should be independent from an external encoding. Currently the external encoding is duplicated also in the font, unnecessarily, as this info is ignored anyway, due to the full external /Encoding. So we might have similar fonts in the PDF that differ only by their unused encodings. I believe we should keep the original /Encoding inside the font intact (unused anyway, only for information), using the value pairs coming with the font file. We can reduce the size of the /Encoding vector in the font so that only the entries of actually used glyphs are there. E. g. if an external /Encoding says 13 --> /foo and the internal original encoding says 55 --> /foo then if 13 --> /foo is typeset by pdftex, we need to lookup /foo in some hash or tree for the font file so we get the original 55 which goes into the subsetted font file stream. I'll try this. Then there is some magic needed to trigger glyph activation in one PFB file only... Regards, Hartmut