[xtm-wg] Re: parallel development of syntax and concept models

["Luis J. Martinez" <luisjm@luisjm.com>]

[Luis Martinez <luisjm@luisjm.com> on Tue, 29 Aug 2000 01:32:24 -0000]
> --- In xtm-wg@egroups.com, "Steven R. Newcomb" <srn@t...> wrote:
> [Luis:]
>> When you said:
>> 
>> "(1) UML models for the abstract expression of constraints on
>> APIs to an information set (such as the information set
>> of topic maps), and"
>> 
>> I don't think the purpose of the UML conceptual model is to express
>> constraints on an API.

> Maybe not, but it does have that effect.  A UML model represents
> decisions and conclusions about what the object classes are.  Those
> decisions and conclusions constrain APIs that are built to conform
> to them, don't they?

Of course it has that effect. The object model provides a way to promote 
the understanding of what a real world topicmap is, and a practical basis 
for implementation. I don't understand what is your point to relate APIs 
to an abstract object model.

>> I think that the conceptual models are to express the Platonic forms
>> of topicmaps independent of OS, programming language, or platform.

> I agree with you, but I also make the same claim for a set of syntax
> constraints.  True, the use of the DTD formalism implies the use of
> SGML or XML, but the constraints expressed by a DTD could at least
> theoretically apply to other markup syntaxes, as well.  Suppose we
> used sets of BNF linguistic productions instead of DTDs.  Wouldn't it
> then be just as true to say that "the BNF productions are to express
> the Platonic forms of topicmaps independent of OS, programming
> language, or platform"?

>> I know that by definition that is impossible to accomplish. But, at
>> least that is the purpose. I think that is where we disagree. Our
>> views of the usefulness of the conceptual model is very different.

> OK.  I see that we don't agree about this.  I would ask you a
> question: Is there nothing about the Platonic forms that is not
> captured in a UML model of them?  It worries me that you don't see any
> distinction between your preferred expression of the Platonic forms (a
> set of object classes) and the Platonic forms themselves.

I am completely aware of the distinction. I know that anything that represents
the Platonic forms is an imperfect copy. But, if we do not base our 
implementations on an imperfect representation, what are we basing on?
That is what I am not understanding. 

>> Also, i think an object model is more abstract than a DTD. You can
>> use a DTD as a modeling language but it is too ambiguous to define a
>> structure by itself.

> This statement, while true, does not speak to my point.  I'm willing
> to grant that there are certainly ways in which an object model is
> more abstract than a DTD, if you're willing to admit that:

> (1) Both DTDs and sets of object classes are both expressions of a
>     higher-order abstraction (the information set, aka the Platonic
>     forms), and

> (2) DTDs and sets of object classes should be designed with different
>     engineering constraints in mind, because their scopes and fields
>     of application are different.

This is my point. I do not think that an object model scope and field
of application is to just implement Application Program Intefaces. It
is broader than that. Is to help us understand what we are trying to
design here. 

<!ATTLIST
  name
   scope          -- List of references to scoping topics that
                     collectively define the limited contexts within
                     which the name is applicable to the topic. --
      CDATA       -- Reference --
                  -- Reftype: topic+ --
      #IMPLIED    -- Default: scope is unconstrained. --
>

scope CDATA #IMPLIED with a bunch of comments next to it is just not
good enough for me. Prose is great for lawyer but not for programmers.


> I do not claim that one is more or less abstract than the other, and I
> do not claim that one is *not* more or less abstract than the other.
> What I claim is that this whole question of relative abstractness is
> irrelevant, as a practical matter, and extremely misleading, because
> their scopes and fields of application are orthogonal, and because it
> makes it far too easy to succumb to the temptation to ignore the
> distinction between

> (1) how and why something is being said about an idea, and 

> (2) the idea itself.

> This is a really big and really important distinction.  To fail to
> make this distinction is to make the same kind of error that is made
> when one confuses:

> (1) a generally accepted scientific theory

> with

> (2) the truth.

I understand the distinction. If we don't have an explicit, imperfect,
flawed representation of topicmaps, all we have is an implicit model
of topicmaps left to interpretation. Again, that is great for laws but
not for computers.

> Forward progress is greatly hindered, and even prevented, when we fail
> to make these kinds of distinctions.  Making these kinds of
> distinctions allows us to make progress *in spite of* our inability to
> perceive the truth directly.  First of all we must understand that our
> whole perceptual and cognitive apparatus, including our system of
> beliefs and our latent assumptions, are almost certainly flawed.
> Having understood this, we are in a position to adapt appropriately to
> evidence that we've been wrong all along, or that there are depths of
> understanding of which we have been unaware.

So we should not make decisions for fear that we might be wrong? 

> I personally believe that dialogue between the modeling group and the
> syntax group will bring a greater understanding of the basic ideas
> here.  I do not believe that either group's models will be the truth,
> or even that the combination of all models will be the truth.  The
> truth is much more elusive and intangible than that.  But the truth
> *does* exist, it *does* await further unfolding, our apprehension of
> it *will* improve.  That improvement will be greater and sooner by the
> use of reasonable and reasonably diverse approaches to expressing it,
> especially in cases where multiple diverse expressions are *required*.

> -Steve

> --
> Steven R. Newcomb, Consultant
> srn@t...

> voice: +1 972 359 8160
> fax:   +1 972 359 0270

> 405 Flagler Court
> Allen, Texas 75013-2821 USA
> --- End forwarded message ---



-------------------------- eGroups Sponsor -------------------------~-~>
GET A NEXTCARD VISA, in 30 seconds!  Get rates
of 2.9% Intro or 9.9% Ongoing APR* and no annual fee!
Apply NOW!
http://click.egroups.com/1/7872/4/_/337252/_/967514946/
---------------------------------------------------------------------_->

To Post a message, send it to:   xtm-wg@eGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@eGroups.com