Re: To gh_ or not to gh_?

[Marius Vollmer]

"Lars J. Aas" <address@hidden> writes:

> On Sat, May 12, 2001 at 04:19:07AM +0200, Marius Vollmer wrote:
> : "Lars J. Aas" <address@hidden> writes:
> : > I come from the school where the thinking is that internal functions
> : > shouldn't be present in public (installed) header files at all.
> : 
> : I wouldn't want to have non-installed header files.  I agree that it
> : would be an improvement to distinguish between different levels of the
> : Guile API, in the sense that features that are more difficult to use
> : right are on lower levels and that users know which level a certain
> : feature belongs to.  But I, at least, don't want to shut them off from
> : the lower levels with technical measures that are difficult to
> : circumvent.
> 
> I assume this is a vote against the SCM_INTERNAL scheme too then?

Err, no.  That would be technical measure that is easy to
circumvent. ;) Private header files that _are_ installed but not
included by default anywhere would also be OK.

I just wouldn't like to have the situation where someone knows that he
needs that particular internal structure definition to patch around
some bug in some releases of Guile, but that structure definition is
not available in the installed image of Guile and he decides to copy
it from the source of Guile into his own program.

That would probably work for any specific version of Guile, but it
does have catastrophic failure modes when the two copies of the struct
definition drift out of sync.