Re: [Texmacs-dev] Darcy's plan for GNU TeXmacs 2.1.1

[TeXmacs]

Dear all,

On Wed, Jan 12, 2022 at 10:55:38AM +0100, Massimiliano Gubinelli wrote:
> - A problem with Guile 3 (and some other options) is that they use proper 
> unicode strings and this clashes with our liberal use of strings as byte 
> buffers. I had to do some hacky stuff to get things going in the impl.

And god knows what else they will come up with to break your code in Guile 4.
Also, Guile 3 is no option as long as it does not support all operating systems 
that
we currently support.

> - I think our scheme codebase does not really depend on any specific 
> implementation and we could try to keep it compatible with Guile 1.6, Guile 
> 1.8, S7 and Guile 3 by havign all the specific code in some (one?) file and 
> keeping all the other generic. Pervasive changes are required for Guile 3 
> (due to the fact that definitions are not possible in conditional code) but 
> the added code could be made no-op with macros in the other implementations.

Yes, whatever solution we opt for, there is no reason to remove your generic 
layer.

> - s7 seems to me a promising option. I think we could discuss with the main 
> developer to solve the issues we see: better symbol lookup in some situations 
> and better memory management. 

Don't forget that our use of Scheme is quite special:
I do not think that we use any cyclic objects (at least as far
as my own code is concerned).  In principle, reference counting with
occasional mark and sweep would probably be the best algorithm for us.

For TeXmacs 2.2, we still need to decide something.  If we go for Guile 1,
then we should further improve the solution that we put into place with Denis
so as to include a minimalistic Guile with the source distribution of TeXmacs.

Best wishes, --Joris