Magnus Berg (qmgb@bull.se)
Mon, 16 Mar 1998 17:37:08 +0100
Hi Doug
I noticed that this post got a little technical, so if anyone wants me
to elaborate, just ask.
> You appear to imply that the moral codes are formal. But in the MoQ the
> moral codes are indeed Static Patterns of Value subject to the dynamic
> force of evolution (to use Ken's, et al., terms).
Yes, the moral codes are of course subject to DQ, hence the "no DQ though",
in the other post. An OO model of static patterns would only be a *static*
model. As soon as interaction gets involved, we leave the MoQ, because
interaction is predictable in the OO model but not in the MoQ.
The cause of the predictability in the OO model might be that we *are*
able to observe it passively from the outside without getting our
noses wet. The entities inside the OO model would not, so it might
seem unpredictable to them, would they ever reflect about it.
> Why not use, e.g., genetic algorithms to emulate DQ? As I referenced
> near the end of 1997 in email here, there are folk doing adaptive,
> self-reprogramming agents in silicon as we speak.
>
> Also, why couldn't the SPoVs be heuristic? The hard part here is
> emulating context, its persistence and its adaptation. Growable
> associative memory plays a large role in the technical solution to this
> problem.
I'd go for the heuristic approach, it implies a certain degree of
unpredictability as opposed to algorithms. Maybe it's pseudo dynamic
enough, maybe not.
> I guess what I am saying is that your Object/SPoV-Oriented approach
> needs a hybrid platform on which to achieve a more complete emulation of
> MoQ.
You mean a hot cup of tea? :) Don't think I've never thought about it.
> I looked on the amazon.com site and found Gelernter. Wow! He is a true
> futurist! I discovered that his ideas are at least six years ahead of a
> few of my own. I ordered his book, 'Mirror Worlds.' Do you know if he
> is the same Gelernter who was the unabomber's 14th victim?
Certainly hope not, don't know about the unabomber though. His latest
book was published in -98 so he was still kicking quite recently.
> Thanks for your reply above. I was not attempting triteness on the
> issue of (software) Objects as SPoVs. I sense the presence of SOM
> influencing our jargon and that (IMO) may confuse casual visitors.
>
> Also, when I think of the O-O development paradigm illuminated by MoQ's
> SPoV-O light the focus moves from encapsulating methods/resources in
> Objects to aggregating context-sensitive (a la many truths) resource
> behaviors in interrelationship value patterns. (Admittedly this is
> easier to say than do, and bio-components' quantum interrelationships
> facilitate this well in that realm. It is tougher to accomplish in
> formal code on a formal von Neumann architecture machine. Now there is
> an entrepreneurial opportunity for someone!)
Actually, what the MoQ says about this is that modeling an interactive
software inorganic level would suffice. Other levels would appear in
time IF it's dynamic enough. The OO approach would be to cheat and
provide a ready made, maybe too static, framework.
There's a more thorough discussion about this in my classicist essay
in the forum.
Maybe what is most appealing about the OOMoQ idea is that it might be
a good way to design systems. Simply get rid of the popular three tier
design method and rip off the four layers of the MoQ instead. As you
say, it's an entrepreneurial opportunity all right.
> MoQ shows us that is where the real leverage on Value is. Contemporary
> OO technique (I am NOT up to speed on the latest techniques!) appears
> still trapped in an intra-object resource aggregation camp. My guess is
> this is primarily influence from some of DeMarco's and Yourdon's old
> ideas on functional cohesion. Unsure you have read them, and from your
> words I sense you are a student of Booch.
>
> Imagine what our cells would look like if our life forms were designed
> with cohesive intra-object resource aggregation. They are not,
> however. All cells are virtually (method/resource-) identical, with
> each carrying the entire specification for the final composite form.
> The homeobox interrelationships among the cells tell each which tissue
> to build dependent on context. That is NOT functional cohesion! It IS
> re-use or design cohesion.
>
> By comparison functional cohesion would demand unique cells for each of
> the roughly 100k proteins which compose a modern human. To me, this is
> the way we do it today. A better model is the one described in the
> previous paragraph. BTW the 'better' model is demonstrably more MoQ and
> less SOM. :)
You are right, I fear, that the most common way of designing OO systems is
what you call functional cohesion. I didn't recognize the term but I get
the picture by your examples.
I'm quite reluctant to actually call that OO. There are way too many
so called OO tools and languages that call themselves OO just because
they use the OO dot notation. They are often very crippled in terms of
the more useful OO features such as polymorphism, multiple inheritance
and operator overloading.
To me, that's a real problem sometimes. People in this business are
too concerned with making systems simple, and simple according to
them means procedural and makes it simple to follow the execution
of a program. It's a flat, one level design paradigm that might be
simple, as in brainless, but it has none of that power associated
with beauty in Gelernter's latest book "Machine beauty".
Magnus
-- "I'm so full of what is right, I can't see what is good" N. Peart - Rush-- post message - mailto:lilasqd@hkg.com unsubscribe/queries - mailto:diana@asiantravel.com homepage - http://www.geocities.com/Athens/Forum/4670
This archive was generated by hypermail 2.0b3 on Thu May 13 1999 - 16:42:56 CEST