Re: gh_enter reconsidered

[Mark Galassi <rosalia@cygnus.com>]

    >> Maybe we should support gh_enter, because it's portable but
    >> intrusive, but also provide less portable and less intrusive
    >> alternatives.

    Klaus> Why are non-intrusive interpreters less-portable? Perl
    Klaus> looks pretty portable.

Perl does not have a conservative GC, so it's not the same issue at
all.

But I think a better wording is that gh_enter() is more robust: the "I
steal your main()" implementation from SCM and Guile has *no*
OS-dependent ifdefs or anything.  It just works on Ataris, Amigas,
VMS, UNIX, and so forth.  And if a new bizarre OS comes up, it will
probably work there too.

As Jim said, we will probably end up with system-specific
top-of-stack-finding routines for all architectures.  Then we will be
portable, but not robust: we will need to do the work from scratch for 
each new architecture or C library version or whatever.

I think it's worth the price.