At 13.03 +0200 2001-05-21, Marcel Oliver wrote: >Lars =?iso-8859-1?Q?Hellstr=F6m?= writes: > > My suggestion is rather that the current set of math symbol > > commands should be made into this layer (hence some commands which > > now produce different results would then by default produce > > identical results). > >Could you define the _problem_ you are trying to solve? Good question! On one side, the problem is that in LaTeX today, I'm not expected to write what I mean in math, I'm expected to specify the visual expression for what I mean. This goes very much against the general trend in the development of LaTeX, which is that you should say what you mean and leave to the style (documentclass, packages used, preamble declarations, etc.) to sort out what is the visual expression for this. To follow a principle of "writing what you mean" should furthermore _strengthen_ the role of LaTeX as a standard for expressing mathematical formulae in ASCII and the like, since writing what you mean is precisely what you do in this case. I personally think that the \epsilon glyph in cmm looks horrible and use \varepsilon whenever I can, but I still wouldn't write "Let $\varepsilon>0$ be given ..." in an email explaining some mathematical argument, I would write "Let $\epsilon>0$ be given ..." because that is how I think about it and that is how would say it. On the other side, there is the matter of font design. Firstly I think it is a designer's right to conclude that there can be no decent looking, say, lunate epsilon (like CM \epsilon) in a particular design and therefore not include one at all. Secondly it is a matter of time; the usability of a font benefits more from time spent on cutting additional characters or improving the shapes of those already implemented, than it does on time spent cutting yet another glyph variant of an old character simply because CM happened to have that many variants. Now what would a math font designer who cuts only one glyph variant of epsilon do today? Most likely put that epsilon in both slot "0F and slot "22 of the OML-encoded font in the family. This will work fine until that day when comes a document where (in the spirit of Hans Aberg but against better judgement) \epsilon and \varepsilon are used to mean different things---then the typeset result will be wrong but LaTeX doesn't know this and thus cannot give a warning about it. By contrast, there would be a warning in the system I am suggesting. With the default \epsilon and \varepsilon there would be a warning that the manuscript is inconsistent, so maybe the author hasn't noticed that \epsilon and \varepsilon do the same thing. If the author redefine \varepsilon to produce a different glyph variant than \epsilon (e.g. by a command \renewcommand{\varepsilon}{\MathUseVariant{\epsilon}{curly}} or something similar) then all is as today with CM, but when using the font with only one epsilon there will be a warning that the requested glyph variant does not exist. >(If you think >the problem is that people write unreadable math papers, I tend to >agree, but there is not much that LaTeX could do about that...) At least about this aspect of it we can do something. The guidance given by LaTeX manuals and the like on what is sensible and what is not is, in this area, rather important. Lars Hellström