On 21/06/2011, Taco Hoekwater <[email protected]> wrote: > > On 21 jun. 2011, at 18:30, Ulrike Fischer <[email protected]> wrote: > >> >> tex4ht doesn't use only this htf-files. At first it loads the tfm's >> of the fonts mentioned in the dvi (I don't know exactly why). > > If it really needs TFMs, I do not think it can ever be compatible with > luatex (or xetex, for that matter). >
afaik, tex4ht doesn't load the tfms to generate the dvi: that's required by some tex engine (pdfTeX, afaik) which generates a dvi with some tailored suited macros (it makes three passes, to ensure the bbl, idx and other stuff are properly compiled); then, in the second stage, tex4ht deciphers the dvi with the htf files to generate some xml files; and in the third stage, t4ht assembles the final html/odt file with the xmls, the images, and all the other stuff generated by the previous steps. I collect that from reading the oolatex script, which actually controls the whole process. So, in my not very enlightened opinion, htf files are necessary only because the tfm files that generate a dvi may have different encodings, so the resulting dvi files are spaghetti encoded, and there is some need to ensure that appropriate utf8 sequences are produced from the messy dvi into the the generated xml files. htf files are mainly maps from 8 bit encoded fonts into utf8. If a TeX engine could read and write files properly UTF8 encoded, the need for htf files would be bypassed; tex4ht would only have to translate typesetting instructions (from a target successor of dvi format) into xml tags, since the encoding would be UTF8 right from the beginning. -- Luis Rivera O< http://www.asciiribbon.org/ campaign
