Seems FFI is the main blocker for Slate now, right?

[Ken Causey]

On Sun, 2004-10-10 at 09:58, Pupeno wrote:
> On Sunday October 10 2004 07:01, Paul Dufresne wrote:
> > Real Slate applications depends on a GUI (Graphical user interface).
> > Is FFI waiting for something too?

> Yes, a developer ;)

As Pupeno implies below I've done a little work on this.  But Pupeno is
right here as well.  I've only barely scratched the surface in a
prototype and Fall tends to be a busy time for me so I can't make any
promises that I will make much progress in the near future.

> 
> Can anybody tell me some docs I could read to understand FFI ? Ken Causey, I 
> know you've been working in FFI, can you tell us what you've done, and can 
> you share it with us ?
> 

Well I can give you the links that others gave me:

http://xarch.tu-graz.ac.at/autocad/lisp/ffis.html (Unfortunately this
one seems to be closed off now, perhaps that's just temporary.)

http://common-lisp.net/project/cmucl/doc/cmu-user/aliens.html

And then I looked at Squeak's use and implementation of FFI somewhat.

What I have done so far is only really the minimum to test my
understanding of the use of dlopen/dlsym/dlclose and libffi.  What I
have done will only work on systems that support these, certainly not
including win32.  Right now there is no real library and simply as
single method that allows you to call a function of no arguments that
returns an integer.  Not particularly useful.

If anyone thinks they would still like to see what little I have done
then let me know and I will get it together.

Ken

P.S. Let me state something that I think is more important to the
current discussion:  I see no reason that anyone should be waiting for
FFI.  I think that even when it is completed FFI is a costly interface
and should not be used for any library where speed is essential.  It is
useful perhaps for quick prototypes and the occasional library call
where a lot of time is spent in the call and the overhead of calling the
function is not all that relevant.  That is not the case for a GUI
library.  I'm of the opinion that such a library MUST be implemented as
primitives as it simply would not be useful implemented via FFI.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : /archives/slate/attachments/20041010/dc60508d/attachment.pgp