I have put an updated version of my experimental math font implementation at the address posted earlier in this thread. The fonts are more complete now and the encodings have been reorganized a bit. The most important deviation from Jusin Zieglers proposals is that I have split the encoding containing TeX-specific characters in two parts, MX1 and MX2. I have written a note on why I think this is a good idea; see below. Comments are welcome and, in fact, needed. Matthias Splitting MX ? While in general the proposals given in the `Technical Report on math font encoding' (J. Ziegler) are very sound, there is one place where a deviation might have big advantages. After having seen the yhmath package by Yannis Haralambous, I think that the role of the MX encoding is a bit different from the other proposed math encoding. In MX many characters are not accessed directly, but just through charlists or extensible recipes. This gives the font designer quite a bit flexibility in deciding e.g. how many big variants of a delimiter or an accents to put in the font. This does not affect the macro programmer in any way, as long as the entry points to the charlists are in the same slots and at least as many sizes as in cmex are given (for compatibility). But this flexibility can only be exploited if there is enough room in the encoding. The MX proposed in the cited paper leaves very little room (every delimiter has room for four sizes, the accents have room for eight sizes), since it includes many new delimiters and operators in an attempt to collect all TeX-specific glyphs in one font. This goal is not achieved anyway, since some wide accents have to live in other symbol fonts. The new delimiters and operators in MX do not have to live in the first 4 (or 6) families, since they are neither in plain LaTeX nor in AMSLaTeX. So one might argue that anybody who wants to use these glyphs has to use another (fifth/seventh) family anyway, so why not put them in an additional MX2 encoding containing more TeX-specific glyphs ? If the new big delimiters are no longer in the first 6 families, there is not much point in keeping their basic size counterparts in MC. These should also go to MX2, freeing some valuable slots in that encoding. I think the splitting of MX in MX1 and MX2 has many advantages: - giving all users of plain LaTeX or AMSLaTeX improved quality of big delimiters and wide accents (since there is enough room to provide more sizes, a la yhmath). - giving users who need the new delimiters or wide accents improved quality compared to the MX-proposal (since there is enough room in MX2), while not forcing them to load more than one additional family, which they would have to do anyway with current LaTeX or AMSLaTeX, so that there won't be compatibility problems. - freeing some valuable slots in MC. These could e.g. be used to turn the group and moustache delimiters into `first-class delimiters' by giving them basic size counterparts. - moving *all* charlist-accessed glyphs (even the wide accents) out of MC, MSP, MS1 and MS2 . - there is now enough room in MX2 to add some more wide accents, like wide triangles or parens (found in yhmath). I have updated my implementation skeleton to reflect these thoughts. Are there any serious disadvantages, I have overlooked ?