>>>>> "Asger" == Asger K Alstrup Nielsen <[EMAIL PROTECTED]> writes:
>> I think we have to choose between a small and relatively
>> cost-efficient modification of the current format or a full blown
>> guile format. The current textclass format is really an ad hoc one
>> (entries are not orthogonal at all) and we cannot afford to add new
>> functionality if it does not have clearly defined semantics. We
>> should be able to differentiate for example between the code
>> generation part and the display part. Currently, several options
>> serve both purpose, and in a way which is sometimes ugly.
Asger> Maybe the first thing to do is to understand exactly what is
Asger> wrong with the current format. Which entries are not
Asger> orthogonal? Can they be made orthogonal? Can we simplify
Asger> things? Can we separate things?
For example, you know that a description entry should be output as
\item[firstword] secondword
because the LabelType is 'manual', whereas LabelType should be a
display-only thing. Similarly, the bibliography has a labeltype
bibiliography and that's the only way to recognize it.
There are many more examples when only reading the code can help to to
guess how things work. It seems that these variables were added when
people needed them, without really trying to get things right.
Counters should probably be defined in textclasses as different
entities too, so that we can have some for theroems and friends.
Asger> In short: Can we clarify the semantics of the format, and then
Asger> simplify the semantics of the format, while still extending it?
I think we should try to do that. A it of cleanup is necessary.
Asger> Given that the current format is an ad hoc solution, chances
Asger> are that this task is possible without doing a full-blown Guile
Asger> format.
It depends. Look at the other thread about SGML...
Asger> I fear adding a full-blown Guile format (btw. guile 1.3 is out!
Asger> :-) will lead us into a trap: We'll just get even worse
Asger> semantics and thus more bugs.
I was just telling that to avoid the temptation of a more powerful but
bloated format. I do not think it would be wise to introduce guile
before 1.2, anyway.
JMarc
