Actually all I wanted to do today regarding LaTeX was replying to Javier's mail. instead ... Anway, here we are now (still before midnight) ... > > it seems to me that this all boils down to "i want to ensure that all is > > Type1" so that i get proper pdf files. or am i wrong? > > That could be a possible reason, but very likely some people could > find other reasons. But yes, that is what I was thinking about. is there _anybody else_ who can come up with another reason why one wants to restrict a list of possible encodings (of the same font, or say set of glyphs!) other than some of the actual fonts are not suitable for the target device, eg pdf output? if so please tell us on the list. > More generally, the automatic selection should be done taking > into account possible restrictions/features defined by the user. > (Which restrictions/features can be set is to be studied.) it is certainly true that such wanted restrictions/features should be studied but can you give us any clue what you are thinking about? having looked a bit at the current internals of NFSS2 (my God last i really did this was long long time ago) i can see some easy way to attach such restrictions on the level of either a) the font family b) the encoding/fontfamily combination but i'm not sure one wants to look at it in this way. only what could be the way to look at it? > It should be noted, however, that text also contains symbols and that > usually we are using _two_ encodings -- if we say \defaultencoding{T1}, > \textdagger is taken from cmsy, which is the default symbol font (OMS). > So we have another problem when selecting an encoding -- in fact > we must select two. do we? i'm not so sure. things like \textdagger aren't really tight to a language, are they? i mean within one font you will have not different variants to be selected depending on the text being French or German or so. at least i would consider this extremely unlikely and would think a language model should not explicitly build for it. however you are right, that there might be the problem that for some fonts you have TS1 available and for others you don't and you might want to mix the fonts. in the latter case things get more difficult. > Instead of multilingual documents, I would rather speak of multiscript > documents, since imo this concept is more important. Of course, different > languages use the latin script in different ways, but when there are > languages using diffent scripts there are additional problems. For > example, in a text containing both latin and cyrillic scripts the > following line of code might not make sense: > > \usepackage{times} > \usepackage{textcomp} > > Is times applied to both scripts? not sure what you mean by applied to both scripts? the "times" or the "textcomp"? > What happens if for some reason > I like cyrillic times and latin palatino? first of all that needs to be specifiable and that is an interesting problem in itself (though solvable in a decent way). it is not quite clear which atributes should be modified from language then. i guess the right approach is to consider all commands that are described on p173 of the LaTeX Companion (eg \famildefault ... \updefault) as language dependent commands but it is kind of tricky to find the right level for specification and behaviour > Which symbol fonts are used? well TS1ptm for cyrillic and TS1ppl for latin > And if there is no times font for cyrillic? too bad then you shouldn't request it (and you get the default cyrillic (eg default T2A font) which might be some cymr font) > Or there is no symbol > font? again, you get the default TS1 font. but there is not much you can do about that since if you ask for, say \textyen you better select a font actually containing it, don't you? (okay, i'm simplifying but it is getting late now :-) > If you use latin with tibetan, things are more complicated > because tibetan doesn't understand what a times font is, and again, ??? you mean, there is no tibetan Times glyphs/font? --- covered I think by the posibility to select \rmdefault etc on a per language basis and by the fact that for each encoding there is a default font which you get if things fail. > what happens if a certain script requires two or more metrics and/or > encodings? i would be grateful if you elaborate on the last two with examples. > Finally, it remains the problem of what happens if people dislike > the default settings. But that is unavoidable... it is a real problem if you forget to build your system in a way that the defaults can be easily altered. if they can it might still be a problem, but then i vote for, too bad (if you honnestly think you have chosen good defaults:-) really good night now frank