Eliot Miranda wrote:
>
>
> On Sat, Dec 5, 2009 at 2:00 AM, Stéphane Ducasse
> <
[hidden email] <mailto:
[hidden email]>> wrote:
>
>
http://lambda-the-ultimate.org/node/3702>
>
>
> I want to return to the implementation and optimization implications
> of this later. But Steele's post refers to William Cook's excellent
> paper On understanding data abstraction, revisited. Onward! Essays,
> 2009 ACM OOPSLA. SIGPLAN Notices 44, 10 (Oct. 2009), 557-572. There's
> a freely available draft version
> <
http://www.cs.utexas.edu/%7Ewcook/Drafts/2009/essay.pdf>. The quote
> I want to emphasise is
>
> The extensibility of objects does not depend upon inheritance, but
> rather is an inherent property of object interfaces.
>
>
> The paper proves this point eloquently and concisely. The implication
> is that isKindOf: is <shout>*/BROKEN</shout>/*. In the absence of
> something better isSomeInterface to query whether an object implements
> some interface SomeInterface is acceptable. isKindOf: (and the horror
> I found the other day isKindOf:orKindOf: (you see why it exists and
> where it leads?)) is not and should be expunged.
>
> My weekend rant.
>
> best,
> Eliot
I fully agree with #isKindOf: and the reasons given in the paper.
But in my opinion, the #is: method as suggested by Igor at
http://lists.squeakfoundation.org/pipermail/squeak-dev/2009-June/136793.htm
(and included in Cuis) is much better than #isSomeInterface. It has two
big advantages:
- It does not bloat the classes, especially Object, that would need to
implement many #isSomeInterface variants.
- By just looking at the inheritance of a certain #is: implementor you
can get a quick grasp of the interfaces it conforms.
Cheers,
Juan Vuletich
Locked
