Re: Guile FFI: resizable vector problem

Tom Lord <lord-1tv/[email protected]> writes:

> I'm not against the feature-macro idea.  It makes me a _little_
> nervous just because I don't like #ifdef'ed code -- but on the other
> hand, ASCII-only (or iso8859-*-only) optoins would be very winning
> things to provide.

My greatest concern with #ifdef'ed code right now is the issues it
might bring up for the portable binary idea (I still like this idea
for a few reasons, but maybe I'll have to grow out of that later).
Among other solutions, I can think of at least two:

One is that extra functionality that may be present in an
implementation should be included in the standard, but allowed to
return error codes and conforming codes must accomodate for the
possibility of these errors.

The other (inspired by OpenGL) is to provide a function that returns a
string with the names of all the extensions they provide.  Client code
can request this string, search it for specific features, and if
they're available call another function and pass the features name to
return a pointer to the function.  (The user's required to cast the
function pointer appropriately.)  You can think of this as roughly
equivalent to srfi-0 or srfi-7.

-jivera