> > The structure of the document in its first stages is indicated > > with punctuation marks, and typesetting indications such as > > \smallskip, \hspace{2cm}, \bf or whatever extend the set of > > punctuation marks available, and provide a convenient visual way > > to refine the structure. This mimicks exactly what one does when > > writing an article by hand. > um, speak for yourself. I have tried to train myself *not* to write > articles simply as a mimic of how i used to write by hand. on the > contrary, when i write by hand these days i put in `generic markup' You have cut this citation from its context. I did not say that I never use structured commands and use only \hspace{2cm}. I said that in the early stages of _creating_ a scientific article, it is totally unproductive to try to separate structure from formatting, because the structure is largely undefined: when one does research work, one does not know the final results _by definition_ so one knows still much less the final structure of the document. If the history of a scientific article would be done from conception to publishing, I would say it starts as a pure Plain TeX document, as there are mostly mathematical formulas plus some unstructured comments explaining what the formulas are about, separated by commas, points and blank lines. That's all. Then, during the refinement process, as mathematical / physical / whatever new results are found and accumulate, the document evolves toward a more and more structured document, either with LaTeX or with Plain TeX/AMSTeX and personal macros (for many mathematicians). Ultimately, when the document reaches _you_the_publisher_ in its definitive form, then it may be transformed into SGML or anything else you like (PostScript or PDF?). But trying to write a scientific article from the scratch directly into SGML would be absurd, in my opinion. A secretary, maybe, but not a researcher. I hope that only those who never used \noindent or \medskip in an article to tune the aspect of the output, instead of modifying the class of their document (or the DTD for SGML addicts) will throw me some electronic stones at the head :-) Michel Lavaud ([log in to unmask])