Re: But is it Object-Oriented?

>>There is not less data abstraction when you have multi-method than when 
>>you are restricted to mono-methods. In both cases, outside of the 
>>package defining a class, you can only use the operations that are 
>>public.
>>    
>>
>
>The paper that Isaac cited argued exactly against this. The argument cited was
>like this: Some "fundamental" methods need access to the implementation (fields)
>of a "class" and some methods can be implemented purely in terms of those
>fundamental methods. But, how do you distinguish which methods are fundemental
>(privileges to access private members) and which aren't? Let's say that if a
>method implementation has a parameter specialized on a particular "class" then
>it is given privileged access. Otherwise, it isn't. Then, we can recognize all
>the fundemental operations of each "class" by looking at the method argument
>specializations.
>  
>
Yes, this rule breaks data-abstraction (note that I never advocated this 
behaviour). But since this rule is arbitrary, I don't think this shows 
that multi-methods lead to less data-abstraction. It just shows that 
this is a wrong rule, and that you have to use another one :-)

>I think everybody here agrees that this isn't a good idea. Like I mentioned
>before, I think that ADT = module = types+functions like in OCaml/SML.
>  
>
I agree with that. That's the idea for Nice: let the package system take 
care of visibility.

Daniel



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click