Re: Unified name properties

[Dirk Herrmann]

On 2 Oct 2000, Marius Vollmer wrote:

> Keisuke Nishida <address@hidden> writes:
> 
> > How about creating a C function scm_object_name and a Scheme procedure
> > object-name, and using them for every name request?  It doesn't matter
> > how to implement it now.  We can modify it later on.
> 
> Yes, that would be good.

I agree that we should just provide a simple API for names, such that we
are free to choose our implementations.  But, the question about whether
there is just a single name or whether there may be several names is still
open, and it influences the definition of the API.

With multiple names, we should for example name the functions 
scm_object_names and object-names.

My suggestion is, that we allow for multiple names returned in a list, but
that we also guarantee that the first name listed is somewhat 'basic'.  
For example, when using the environment-scanning approach to names, a
search order could be provided which defaults to scanning the r5rs module
first, followed by guile's internal modules and finally user-modules.  
This is most flexible, as it allows to build the `there is at most one
name per object' implementation on top of it.

But, I have the impression that I am the only one who thinks that there
may be several names for the same object...

Best regards
Dirk