[ocaml-platform] Changes to my previous proposal for namespaces

[Gabriel Scherer]

I think this whole discussion could be avoided if we took the approach
of specifying the command-line options and hardcoded conventions as
building one environment. Ordering issues, conflicts possibilities
etc. would all be explicit by the description of how this initial
environment is specified.

> I'm more concerned with the fact that order matters even if there is no
> conflict, which, again is a departure from today's situation.  ocamlfind
> will need to pass -I flags in the correct order.

I'm surprised: isn't the order of -I flags already meaningful in case
of two compilation units having the same name? It looks like this is
the exact same problem (with the difference that a flexible namespace
language could allow users to avoid any such conflict by redefining
conflict-free names in any situation).

On Tue, Mar 19, 2013 at 1:14 PM, Alain Frisch <alain.frisch at lexifi.com> wrote:
> On 03/19/2013 12:58 PM, Leo White wrote:
>>
>> Sorry, the above psuedo-code is supposed to be going backwards through
>> the paths.
>
>
> It should be noted that this would change the current behavior (-I directory
> are scanned in the order they are provided, with a first-match policy).  We
> probably don't care too much because it is unlikely that the same unit will
> be present in several directories.
>
> I'm more concerned with the fact that order matters even if there is no
> conflict, which, again is a departure from today's situation.  ocamlfind
> will need to pass -I flags in the correct order.  Wouldn't it be better to
> resolve aliases in the same way as references in the source code (i.e. using
> all the -I flags)?
>
> -- Alain
>
> _______________________________________________
> Platform mailing list
> Platform at lists.ocaml.org
> http://lists.ocaml.org/listinfo/platform