Rasmus <ras...@gmx.us> writes:
> It's tough. I've /never/ used neither wrapfig nor longtable. From a
> totally subjective point-of-view I'd certainly want to remove it!
> However, I wonder if this is the 'nicest' thing to do. Not everyone
> cares about LaTeX and not everyone cares to look into LaTeX details.
>
> Three possibilities are
>
> - Just Workᵀᴹ :: Include a lot of stuff in
> `org-latex-default-packages-alist'. Self-proclaimed 'power
> users' can cut it down themselves in their config. It could
> slow down compilation, especially if policy is too lenient.
> (E.g. to support tikz files you need to load TiKZ; To
> support #+LANGUAGE you need to load babel). Perhaps we could
> add an optional variable org-latex-load-all-relevant-packages
> that loads all known packages that Org might depend on (assuming
> they are all compatible). People with i7 processors can then
> turn it on and we could include only basic package in the
> default package alist.
>
> - RTM :: Be better at documenting when a feature requires an
> additional package. This is probably my preferred
> solution.
>
> I think Org can mostly guess when a LaTeX export failed.
> If so, perhaps we could be give informative hints when
> something fails. E.g. if rotation is required and
> something fails, tell the user that the rotation package is
> needed. I have no idea how much work this would be.
>
> - Do nothing :: People who use the LaTeX exporter should be
> proficient enough with LaTeX and Org to solve their
> own problems.
>
> On Eric's original idea about auto-including packages: I don't like.
> I want to like it, but it's just too fragile. Some things depend on
> being loaded in the correct order (e.g. hyperref needs to be towards
> the end). Since people can load arbitrary code using #+LATEX_HEADER:
> \input{·} it's bound to break!
I'm not talking about auto-including packages, nor do I suggest to
include lot of stuff in `org-latex-default-packages-alist'.
My point is: if "wrapfig" is there, "rotating" should accompany it, or
both should be removed from the variable. Also, there's no reason for
"longtable" to be included.
IOW, I'm discussing a very practical point, not the general status of
packages in the LaTeX export back-end.
Regards,
--
Nicolas Goaziou
