Re: To gh_ or not to gh_?

[Rob Browning]

Sam Tregar <address@hidden> writes:

> Because RMS said so?  Using google searches to follow broken links
> from the Guile docs I recently read the "TCL wars" archives.  RMS
> uses the language-angnosticism of his proposed GNU scripting
> language as a major retorical point justifying a new language.  If
> you read his statements literally it sounds as though Guile should
> be handling TCL, C and Python by now (Perl was also mentioned, but
> that's pure fantasy)...  Has this goal been discarded?  Postponed?

Perhaps it's best to say it's been "delayed" :>  Though recently,
Thomas (I believe) has popped up and is planning on working on the
translators this summer.

Note that I believe the current primary tactic is for the other
languages to be translated via a front-end into a low-level scheme
representation that will then be executed by the interpreter, a VM, a
JIT compiler, or perhaps .so code generated via C->gcc->so
conversion.  All of these options are being discussed on guile-devel
right now, and we may end up with a hybrid in the end.

Note that unless I'm misunderstanding you, this probably doesn't have
a lot to do with the gh_ interface.  The "interpreter independence"
claim for the gh_ interface, I believe meant independence from any
particular quirks of Aubrey Jaffer's SCM implementation of Scheme from
which Guile's interpreter originally came.  So you want to have an
abstract:

  SCM gh_add_numbers(SCM x, SCM y);

rather than something like (note that this doesn't actuall exist):

  SCM_BIG64NUM scm_add_big64num_to_fixnum(SCM_BIG64NUM, SCM_FIXNUM);

etc.

-- 
Rob Browning <address@hidden> PGP=E80E0D04F521A094 532B97F5D64E3930