Re: method names in 1.9

On Nov 11, 2007 6:37 PM, M. Edward (Ed) Borasky <[e-mail]> wrote:
>
> Rick DeNatale wrote:
> > Ruby, like Smalltalk, and Self, are languages which exemplify what
> > Ralph Johnson calls the "mystical view" of object-orientation, as
> > opposed the the "software engineering view" taken by C++/Java/Dylan..
> > http://www.cincomsmalltalk.com/userblogs/ralph/blogView?showComments=true&printTitle=Erlang,_the_next_Java&entry=3364027251
> >
> > Quoting Ralph: "The mystical view is that an OO system is one that is
> > built out of objects that communicate by sending messages to each
> > other, and computation is the messages flying from object to object.
> > The software engineering view is that an OO system is one that
> > supports data abstraction, polymorphism by late-binding of function
> > calls, and inheritance."
> >
> > A key difference between these two views is that in the mystical view,
> > the message name (i.e. the name used to represent a requested
> > operation) is defined independently of any type/class hierarchy.
> > Notice that Ralph's description of the mystical view makes no mention
> > of methods or functions. Although in most mystical languages most
> > messages result in the execution of a method with the name of the
> > message, this isn't always true, e.g. when a message is handled via
> > method_missing.
> >
> > The mystical view puts a wall of encapsulation between sender and
> > responder which isn't there in the software engineering view.  The
> > sender names a request, and the receiver responds.  The sender has no
> > knowledge of (and therefore no dependency on) just how the receiver
> > decides to implement the response.
> >
> > Adherents of the software engineering view tend to think, whether
> > consciously or unconsciously, in terms of late-bound function calls.
> > The model is that the caller knows how to find a particular 'virtual
> > function' and then invokes it.
> >
> > It's a subtle shift in viewpoint but it's crucial.
>
>
> Interesting distinction ... but some questions ....
>
> 1. Isn't there a continuum between the two "extremes"?

As a designer, and student of object oriented languages over 25 or so
years now. I can't say I've ever seen a language which mixes the
mystical and software engineering views of object orientation.

In fact, I'm pretty convinced that the two view are irreconcilable at
their core, although many are fooled into trying to bend one view into
the other, usually by trying to force a mystical language like Ruby
into a software engineering view straightjacket.  Chicken Typing is
one product of such attempts.

What I'm talking about here is the language's approach to objects.
Most mystical object-oriented languages are uniformly object oriented
in that everything is an object, most software engineering viewpoint
languages are not, either having objects as an added family of types
(C++) or non-object types in addition to objects (Java).

There are some non-uniformly object-oriented languages which add
mystical viewpoint objects to a non-object-oriented language.  The
best know of these is probably Objective-C.  Another was ClassC which
I designed back in 1982 or so, and never got outside of IBM.  Brad Cox
and I were both inspired by the special issue of Byte magazine which
introduced Smalltalk to the world to add Smalltalk like objects to ,
as a "poor mans" Smalltalk.

> 2. If so, where are Ruby, Scheme, Erlang and Forth on the continuum?

As I said, Ruby is mystical.  Ralph Johnson who has thought more about
objects in Erlang than I have says that Erlang is mystical.

I'm not familiar with standard object-oriented extensions to scheme or forth.

On the other hand, as I remember CLOS and Dylan take the software
engineering view.  I'm more sure about Dylan since I just pulled down
the copy of the Dylan manual David Moon gave me when I visited him at
the old Apple Cambridge (Mass) research lab many years ago, which
explicitly states that Dylan does not do Smalltalk like method
dispatching.

> 3. If either extreme is (or can be made) Turing-complete, why do nearly
> all languages include concepts catering to both?

They don't (see the above).

Plus paraphrasing Matz in his keynote last weekend at RubyConf last
weekend, "Turing complete, Schmuring complete." <G>

-- 
Rick DeNatale

My blog on Ruby
http://talklikeaduck.denhaven2.com/