Of course I read.
Theses do not have to be converted in LaTeX, so there is no such problem in that case. One usually submit directly the PDF.
For journal papers I do not care about numbering, as long as it is correct. Anyway the final version of the papers will have to be composed in the style of the journal, so the numbering can change again. I saw new bug reports, and some indeed seems to be bugs. Of course one has to check the output, we should improve the conversion and have warning when TeXmacs cannot produce a sensible LaTeX conversion. I should say that I do not use myself any fancy feature if I know that I need to convert in LaTeX afterwards, in particular no ornamented stuff, etc… In cases where I need fancy constructions I have TeXmacs macros for which I write by hand LaTeX variants. Anyway I think Joris is happy to have a large repertory of files on which to check conversion. I will also give a look at your bug reports to try to fix something.
An idea for collaborative work with LaTeX users
Remember anyway that TeXmacs->LaTeX is not designed to be 100% faithful, since otherwise there is no point in developing a different system. I think it should be reasonable to point to a conversion which preserve the meaning and translate idiomatic construction of TeXmacs into reasonable approximations with idiomatic construction of LaTeX. Presentation details like the exact shape of ornamented box, etc… it is difficult to reproduce.
It is not necessary to be 100% faithful, but it might be better to give some warnings when it is far from being faithful. For example, it might be OK that ornamented box is ignored and transformed into normal paragraphs, but the user might be better informed, especially note that the error code of LaTeX is in general difficult to understand or locate (this case is understandable, but I think that TeXmacs could predict it before conversion).
For me, it is very important to keep the numberings. We refer to theorems by numbers. At least, I should ensure that the preprint I uploaded, say, on my homepage is the same as the arXiv version. I hope that any published version would be also with the same numbering (if not, the current working style seems dangerous: we usually download the preprints, read and refer to. It is not always the case that we download the official published to check, or even buy it if the library does not subscribe that. In our work, a precise number is very important: nobody will know what theorem is referred to if the author writes “by a theorem in EGA”, say).
I agree with you for the errors. For the numberings I meant the following: anyway the final numbering in the journal will not be the same as in your preprint (if you do not use the style of the journal right away). If numbering is important to you then just make available the PDF generated by LaTeX conversion. More faithful numbering in TeXmacs should be available if you use the same style in TeXmacs as in the LaTeX. If e.g. in TeXmacs you use the amsart style and then in the LaTeX conversion the numbering is different then this is a bug and should be reported. If you use the generic “article” style in TeXmacs, this is not supposed to be equivalent to the LaTeX article style (I think).