Re: The GH interface. (was: Patch for gh.h)

[Rob Browning]

Martin Grabmueller <address@hidden> writes:

> If we wanted people to use GH, we would have to improve it a lot, but
> that would result in a second SCM interface.  That would duplicate a
> lot of code.  Also, I think that a lot of the SCM interface is pretty
> stable too, so using proceedures like scm_make_string or scm_cons is
> safe (wrt future compatibility) anyway.

And we can always (as someone mentioned) have a way of marking public
vs private bits.

However, I will say, after talking with Bill here for a bit, that the
*biggest* problem with the gh_ interface is probably just it's lack of
a clear mission statement.  i.e. Why does it exist, and what is it
for?

  * Is it the preferred interface?

  * Is it the interface for all functions that are C oriented (i.e. have
    C values as args/return values).

  * Is it the "public interface"?

  etc.

I think if we answer this question clearly, and document it, the rest
of the work needed will be fairly obvious.

Even if we just decided that the gh_ interface was the "official" one
and started adding a bunch of #defines to add the stable/useful scm_
functions to it, that would IMO be better than what we have now,
though, I'm not saying that's my preferred or the optimal solution.

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