Good your already on top of this , thought so but wanted to be sure.

> Ok so what is the story here ,,  I  think objects , interfaces and modules
>> all doing encapsulation is 1 layer too many .
>>
>
> Two, actually. :-)
>

Even better.

>
>
>> When you have an interface of statics functions  ( ie a type class or a
>> collection of functions )  , then it begins overlapping the role of modules.
>>
>
> Yes. And I was therefore hopeful that perhaps modules could be replaced by
> static interfaces. If that works, then we can eliminate a concept from the
> language. If not, then we can't.
>
> What I forgot is that module names are tied up in the package namespace,
> which is a different namespace from the identifier namespace.
>

Thats just global space for the package . Why does It need to be that way  ?

>
>
>> Is it viable reverting  modules to namespaces ? Modules are needed with
>> type classes but can  interfaces with namespaces do that for bitc   ?
>>
>
> I don't know why modules are needed for type classes. As to the rest,
> that's the question I'm wondering about too: can interfaces replace
> modules?
>

Are they not need to encapsulate / hide data types and especially their
creation .

In Java / C# interfaces are the contract for modules but its a bit clunky
since you need a reference object .  So you have to create a object which
holds the sub types and methods.    So the holder object and interface in
combination are a module which is not as efficient .

This whole static interface concept makes me go gaga  ( I still have to
respond to the other emails but i need to make a leap ) . Its a contract to
what , ok a contract to functions present in an Assembly ? Across
assemblies ( like type classes)   ?  Or can you use a static class as your
holder ( this is not allowed in C# /Java as interface on static is
meaningless)  which replaces module. Is static class a name for a module  ?

Ben
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to