Being away from home with a mere netbook, I can't really read Eric's and Peter's insightful remarks in detail. I need to get back to a printer before doing them justice. Presentation--at least page size and ease of flipping pages--matters very much. "Semantics only" is pie in the sky.
Consider presenting a little data, which basically comes as a table. Do you print it that way? And if so, do you print it by rows or by columns? Or do you print it as a graph or a bar chart? With bars horizontal or vertical? Portrait or landscape ... Some of these things can be judiciously decided at presentation time, but some the author may really care about. And sometimes the reader may want to play around with it. The latter possibility is in fact the grail of the Unix approach. Data should not be buried in some arcane encoding. Two presentations are in play here. The input presentation should be accessible and processible by unanticipated tools. The output presentation varies not just with medium but with author (and possibly reader) intent. Another place where "semantics only" fails is in maps. Nobody has ever figured out how without multiple encodings of the data to produce a map of a single house lot in the morning and of the global distribution of rainfall in the afternoon. We've heard of mom-only and mm-only users. I happen to be an ms-only user. (Man pages excepted.) A lot of basic style-sheet stuff appears as the value of number registers. There's not really a gaping distance between ms and style sheets--kind of like a small brook that's always been so easy to cross on stepping stones that nobody ever bothered to build a bridge. (Not that the bridge will be easy to build. No doubt the structure underneath the macros will have to be reworked.) Nevertheless, it's a real testimonial that the very first document-level macro package is still a viable tool. Until I get to that printer, Doug
