>>>>> "H" == Hans Aberg <[log in to unmask]> writes:
H> Otherwise, I am not sure it is necessary having all
H> references getting right when working with many files, and
H> doing a subfile compilation. In the scenario I am playing
I myself haven't put much thought in this direction since I was
explicitly trying to write a backwards-compatible system.
There are benefits to the original \include system, but they are not
so great that others should not be considered. The \include system
does not let you do anything you couldn't do with \input. It just
makes it more convenient for long documents. These conveniences don't
seem as wonderful in days of more powerful equipment (it took my high
school math teacher several minutes to TeX one chapter).
In fact there are also pitfalls in the old system that it would be
nice to fill in. As I point out in the "review of the old system"
section of the newclude documentation, it is a convenient feature,
when leaving out, say, chapters 2 and 3 from your book that the
references still work, and the footnote numbers and page numbers of
chapter 4 do not change. This allows me to get output of chapters 1
and 4+ that looks exactly like those sections of the whole document,
without using some sort of post-processor to whittle down the entire
dvi file to the parts I want.
But it is a quite unintuitive and inconvenient consequence of the
implementation that that if you switch the order of chapters 2 and 3
while they are STILL UNINCLUDED, the counters in chapters 4+ are
thrown into chaos.