Re: Recursive namespace

[Keisuke Nishida]

At Tue, 24 Apr 2001 12:50:12 +0900 (JST),
NIIBE Yutaka wrote:
> 
> It is not good to confuse/(or mix intentionally) the name space of
> module and the name space of symbol.  Even if you introduced '::' for
> module syntax, you would be certainly possible to have separate
> differenet name spaces.

That's certainly possible.  I was thinking of a possibility of
unifying those two.

> It is THE PITFALL.  It is quite weird for me, because it is _direct_
> contradiction for _YOUR_ purpose (of compiler).  I have to say, this
> is bad design for compiler.  It introduce nasty things.  If we'd
> implement the idea above correctly, it may result: Module can be
> determined at run-time dynamically.  Say,
> 
>       (let ((module (procedure-which-determine-module)))
>          (module::func blah blah blah))

In this case, the compiler detects that and produces a run-time lookup.
That's also possible.

> See?  I don't think it is good idea.  Module should be determined at
> comile time.  Please think again.  There is no need to use module at
> run time dynamically.  There is no reason to mix the symbol space and
> module space.  The confusion introduce nasty things.

I agree that modules should be determined at compile time.
What I was questioning was whether I really needed to
distinguish module namespaces from symbol namespaces.

> My points are two: 
>       Don't use symbol space to implement module space. 
>       Module should be determined at compile time.

Okay, all systems I know agree with these points.  Probably
there are good reasons to do so.  I'll follow them.