Thierry Bouche writes: > Frank Mittelbach a écrit: > > o should there be user level control on whether or not a heading has > > a number? > > What happens to text like "see section \ref{sec:xx}" if the class > designer withdrawed numbers, the author being not aware of that? that is a question for which there isn't really a good answer. in current LaTeX, the ref refers to the last unit that sets \@currentlabel which means that if in the hierarchy of headings there is a heading which still is number you will get that number. that may not be perfect (eg if that is the chapter but the text says "see section 1" then, but it is better than nothing). some people would like to see the possibility of getting the heading text as a reference (like the IBM manuals do) and come up with "see section ``How to Configure XYZ'', other people say that you can never get this right and definitely can't automatically change from one style to another. it is most certainly the case that any design change of the type "different levels for heading numbers" or "references using heading text" can not really be made without checking and potentially correcting a document afterwards. in other words i think that to some extend the author of a document has to be aware of the layout he/she is targetting for in this respect. (a similar example and perhaps worse is citation styles, eg author/date viz sequence-numbers: you can't typeset a document in both styles using the same source) ================================= i would call two latex classes (the term class is actually bad in 2e) "source compatible" if they can successfully process the same documents, ie provide the same user commands and functionality. in other words a document that can be processed by article.cls would be processable (with sensible results) by any other class that is source compatible to article.cls --- one could say the two LaTeX2e classes belong to the same class, urg (this is when i meant the term was badly chosen). now i wonder if i would claim that article.cls and myarticle.cls differing only in the setting of the secnumdepth counter should be considered incompatible in the above sense. there are arguments for it and arguments against it: - since the availibility of numbering or generally the availability of refering to a heading via \ref differs one can argue the classes are not source compatible - on the other hand this is making the granularity very small and it might mean that coming up with sufficient additional issues of that type nearly no class can be considered compatible to another one. - ... frank ps seems that the post i replied to, took a long time (three days?) to get past the listserv, or was this only to me?