[NTG-context] Global pagination optimisation
j.hagen at xs4all.nl
Sun Nov 22 23:58:28 CET 2020
On 11/22/2020 7:58 PM, Joseph wrote:
> Dear list,
> Out of curiosity.
> Perhaps a question already discussed but wondering if there were plans
> to implement eventually in LMTX an equivalent of framework described in
> article 2018-01-FMi-CI-Journal-28454894_as_submitted.pdf
> or something similar ?
When cooking up solutions I normally start from practical cases, not
from something like discussed in there. I'm more a 'play with something
and see how it works out' than 'read all kind of theoretical stuff' (I
simply don't have the patience for reading up on that). So, context
evolved from use cases and solutions needed. Often solutions then go
through some iterations. It helps that right from the start context was
set up to make extensions relatively easy. Theoretical discussions on
typesetting are wasted on me, sorry. (For the record: I read a lot, on
all kind of topics, but the last time I read something about
typographics is more than a decade ago.)
Now, when discussing document wide optimizations ... I played with that
long ago but kind of stopped when it became clear that it's too easy to
get oscilliation. An interesting suggestion Herman Zaph made when we
discussed this was that like horizontal hz (expansion in tex) one could
also do that vertical: no one will notice it, no matter what folks claim
to notice. That would actually solve quite some issues. It was easy to
make a quick demo then and I might actually bring that into context (if
I can find back the code). Trial and error is my friend (of course with
a dose of thinking and feedback).
It is no big deal to cook up a solution for some subclass of well
defined problems (say: text only with an occasional display element) but
it happens that tex is seldom the candidate for that. Problems show up
when you have plenty of extra elements, backgrounds, floats, notes, etc.
All kind of tricky interactions and constraints. My personal opinion is
that there are no perfect automatic solutions for cases where in fact
the visual appearance matters more: that is why desk top publishing
works ok for complex layouts (it just pays off). It all has to do with
the solution space. Just use the right tool for the job.
Actually, before looking at complex cases it probably makes more sense
to make simple cases look nice and probably the majority of documents
one can find on the web done with tex look quite average. So, from the
context perspective it makes more sense to chellenge users to make nice
documents and provide the means for that, even it occasionally means
some hand work.
Now, with respect to future plans for context / lmtx: I do have some
things on the agenda because there are some half finished experimental
mechanisms in context that could use a bit of help from the engine, but
first I want to finish what we currently have. But, as mentioned: the
starting point is 'user demand' and 'useability' (there are for instance
some things Thomas and I look at occasionally wrt parallel texts that we
need to pick up). And, as with all in context, it starts by just looking
at what is needed, and playing around to get a feeling. Personally it's
more about (visual) fun than about neccessity.
So: feel free to come up with demands but keep it practical and keep
useability and constraints in mind and perspective.
ps. When it comes to columns ... I don't see much of a future in them,
because on devices one doesn't have the constraints of paper: why
complicate ones live. We moved on in other areas too. Actually, if a pdf
viewer could handle a page with height 5 meter well (wrt go back) then
each chapter would be one page and we got rid of floats too. Ok, one
could also use html then -) Of course in magazines, news papers, large
size stuff, separate text flows make sense but that is typically not the
domain where tex makes muich sense.
Hans Hagen | PRAGMA ADE
Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl
More information about the ntg-context