Proposal: module namespaces.

[Malcolm Wallace]

Christian Brolin wrote:

> I think it is better to explicitly distinguish between absolute and
> relative addresses of imported modules. It would be easier for a reader
> (e.g. a compiler) to find the modules. Absolutely imported modules are
> found by looking in the set of roots, while the others are found
> relative to the position of the current module.

I know that relative addressing is important for many practical
tasks.  I'm not yet sufficiently convinced that it is very useful
in module namespaces.  Ok, so actually I can see that it is
extremely useful if modules are commonly going to move from
one part of the namespace to another (the canonical example being
a move from a non-standard hierarchy into a "Std." hierarchy).
However, several people have expressed doubts about whether a
"Std." hierarchy is the correct thing, and no-one except me has
tried to defend the idea.  If we throw away the "Std." hierarchy,
then I'm not sure what compelling reasons remain for relative
module imports?

Regards,
    Malcolm