RE: [Gcl-devel] 2.5.3 is released

[James Amundson]

On Tue, 2003-06-03 at 20:49, Mike Thomas wrote:
> Hi James.
> 
> |I have more
> | thoughts on FFI if there is interest.
> 
> What are your thoughts?

They are not necessarily coherent thoughts, but here goes:

0) I think effort expended integrating foreign libraries into GCL would
be better spent working on a more general FFI.

1) If I were to try to come up with a FFI for GCL, I would probably just
copy an existing interface from one of the other Lisps. I haven't
compared the existing FFI's enough to know which one I would pick. I'm
guessing the ACL interface would be a good choice.

2) Some FFI's seem to be more C-like, some seem to be more Lisp-like.
While I can see arguments for both, it seems to me that the C-like
interface makes more sense because C libraries have C interfaces. There
exist various tools for automatically generating bindings to C
libraries, such as SWIG, that work most naturally with a C-like
interface.

3) In order for an FFI to be truly useful, it has to be symmetric, i.e.,
you must be able call C from Lisp as well as Lisp from C.

4) I think UFFI is a great step forward. HOWEVER,
    a) It isn't symmetric. See (3).
    b) The author has consistently said he would like to make UFFI-2
much better (and different) than UFFI-1. Unfortunately, it seems that we
won't see UFFI-2 any time soon.

I think Maxima really needs a working, symmetric unified FFI that works
with (at least) GCL, CMUCL, Clisp and ACL. The rest of Lisp community
would benefit from such a thing, also.

--Jim