Re: Clean Object Class Design -- Circle/Ellipse

>> Second, Date does not need to know Smalltalk or any other specific
 language
>> to make the points that he makes in his essay on values in general.
>
>Ok. I was addressing a larger point, though. For him to make comments
>about OO, he really should know Smalltalk. If *I*, a mere practitioner
>without a formal computer science background, can read him and conclude
>"he's complaining about things that are easily solved", then there is a
>problem.

Your conclusion falls down here because he does know Smalltalk.

>Yes, how funny. Just what language do you think we should implement his
>approach in?

A new and better one.

>> >I was nodding my head to Date's
>> >points and thinking: "Yup, can do".
>>
>> But at what relative cost?
>
>Today it means giving up the relational model and working with OO instead.
>Of course, this is unacceptable to him.

It is unacceptable to any thinking person who understands the costs of navigational database models.

>> >And that was sad, because by criticizing
>> >C++ and thinking its OO (:-), he gets written off by an field that
>> >could use his help.
>>
>> He did not criticize C++, per se. He merely responded to Stroustrup's
 essay
>> as a widely recognized, respected, published exemplar of the counter
>> argument. He could just as easily have chosen any other published
 exemplar
>> of the counter argument based on any other OO language.
>
>Sorry, I led you astray. I was not refering to Stroustrup. I was refering
 to
>Date's railing against OO where it was plain he was railing against
 *crappy*
>OO languages. Hence my C++ comment.

As far as I can tell, they are all about equally crappy. Why C++ in particular? Why not Smalltalk?

>Now, someone is going to challenge me on what those things/patterns
>were. I have a badly filed scrap of paper in which I wrote it down years
>ago.
>One was (1) So just add the Collection/Relational interface to the
>Object class.

That does not simplify the logical model or empower the DBMS with semantic data or useful identities.

>Another was (2) He really needs to know that its possible
>to program on the class side of the class, e.g., trap every instantiation
>and maintain Dictionaries of all instances of the class.

Who cares? I fail to see what problem that solves. Received on Sun Aug 19 2001 - 08:31:29 CEST