Re: [Texmacs-dev] Towards the TMGUI API

[Joris van der Hoeven]

Let me briefly reply to David's email, which is interesting,
but which goes too far beyond the discussion on the TMGUI API IMHO.

The purpose of the TMGUI API is not to redesign the whole of TeXmacs.
Its purpose is to identify what TeXmacs needs from the GUI and
vice versa. What has to be delegated to GUILE later is not an issue
for this; how boxes are being typeset is not an issue too.
Even embedding extern applications is not really an issue yet;
we may let the TMGUI API evolve for this later.

Improving a big program like TeXmacs in an efficient way implies
choosing a gradual road. When making a profound change like
replacing the GUI, the right approach is to first completely
detach the GUI so that it becomes modular and with a clean interface.
In order to be gradual, it also means that one has to focus
on TeXmacs and the GUI only, even if this what you call
a TeXmacs-centric approach (but really it is not!).

When this is done well, it will in fact not be hard to let TeXmacs
further evolve to what you want. In fact, during the next years,
we should make TeXmacs more and more modular and see each module
as a service which can be used by any other module. Having clean
modules with clean interfaces for all independent parts makes
the whole compatible with many philosophies, just by recombining
the modules in a suitable way. You will be able to embed programs
in TeXmacs, embed TeXmacs in other programs, use TeXmacs as
a typesetter only, use TeXmacs as a blackbox for letting
computer algebra systems communicate, etc., whatever...

So please let us focus on the *minimal* TMGUI API first.
We would like to port TeXmacs to this API using different toolkits
(the existing one and at least one new one) within a reasonable
amount of time. Remind that we should specify a nice, simple
roadmap on how to concretely do the porting in a gradual way;
this is what may allow Stéphane to make some real progress.