Roozbeh, > Javier explained the idea somehow. I'm going into complete explanation > here. Please ask for clarification if there are ambiguities. thanks, I think it has helped, but as you can imagine there are a number of questions. very first one, let's for the moment forget about the high-level NFSS commands like \textbf and so on and just concentrate on the low-level interface which can be perhaps sumarised as fonts are classified according to four different axises NFSS calls them family series shape (and implicitly sizes) but to some extend at least "series" and "shape" are abstract names. first question: assuming you give adequate meaning to these names, are the fonts used in your typography adequately modelled by four axises? from what i understand the answer is probably yes. ================== the series attribute for Latin fonts actually encodes a combination of width and weight; I did choose this for the model because independent changes seemed seldom at least not within a document (i'm talking only about text fonts here not math!) second question: i do understand that you have some characteristics in the fonts that are suitably modelled as a series though you say you consider outline a part of it. is that roughly correct? actually although outlines are mentioned in the LaTeX Companion (and probably in fntguide) as a value for the shape attribute, I think it is something that actually far better belongs to the series if one sticks within 4 axises ==================== > In Persian, we usually do not have the three classic families. In Iran, well, Latin typography doesn't has these three either, i'd say. it is true that we have some understanding what is meant by a typewriter font but this is essentially - by its use as computer input output - by its being monospaced categories for font families is more like 6-9 depending on school, but existance of serifs is clearly an outstanding attribute for a latin based font. However, mixture of \sf and \rm in running text is rather uncommon (unless for very specific items ---the LaTeX commands \textsf and \textrm are really a historical accident due to what is there in plain TeX). i'm not saying that there aren't designs mixing serifed and nonserifed fonts, eg in headings or running heads or other design items. ==================== > Because of this lack of option, outline and shaded-outline shapes are used > much more in technical books, sometimes together with slanted and > backslanted. The model that is used in available Persian software, > modeled from how designers think, is something like this: > > family weight shape > ------ ------ ----- > normal medium upright > italics bold slanted > some others outline backslanted > shaded-outline you say that italics are considered by designers a family. okay, but you don't have all the shape attributes for such a family, do you? ie upright, slanted, and backslanted? anyway, if you just consider Persian text as such (not multilingual documents mixing it with latin, say) then i don't see why the NFSS model would not serve. On the high-level commands you may need something like \itfamily and disable \sffamily and something like \emph would select a family rather than a shape etc but the underlying model would not be affected, would it? perhaps it would be worth thinking about how the high-level interface could be made more adaptable to different languages. ==================== > This the Persian model only. Then you need Latin. For each of the styles > that also may contain Latin text, they should use an equivalent fonts > that goes with it. There are some difficulties here also, one of them you are right, multilingual documents are different. I think the first question that needs to be asked is: Do we have a predominant language A with infrequent embeddings from a different language/script B? In that case I guess one should try to map the meanings of the high-level commands as best as one can using those from the dominant language, eg Something like "bf" for latin scripts could perhaps be translated to mean "outline" in persian context or it could mean "bold nonextended". the idea being that one tries to make the best of \textbf{Some English text with a few \langfragment{persian}{Persian words} in the middle} so that this blends as best as possible (clearly a designer issue). This should work both ways, ie if you have different high-level commands in languages like Persian and you have a multilingual document with dominant language Persian one should map the "persian-high-levels". More complicated perhaps are situations where the different scripts are really equivalent in use. do we then map all such high-level commands, do we use them all, do we just use one set, nevertheless? no good answers here i fear ======================= back to the model of "classical" families > In the absense of that model, designers choose some families (for > a mathematical book, I've seen from as few as one, to as many as six or > seven), and specify that this heading or that caption should come out as > in this family and that shape and size. but this is exactly the way design works here as well. okay, article.cls makes all headings simply by using the body font in a bolden version (and since the default is CM fonts in bx) but classes with a real design would typically specify individual fonts in certain weight, shape and size for the individual elements of the design. what is perhaps missing in LaTeX are different levels of emphasis, ie we have \emph{...} or {\em ...} but for any other textual emphasis people have to use \textbf or \textsf or the like rather than being offered and abstraction which could then be filled with life by the designer. frank