The Role of Universal Thinking In The Human Civilization.

[Siamanto]

The following conversation with axel not only motivated the creation of - but, also, seems as an adequate introduction to - a thread to discuss
1. The nature and meaning of "Universal Thinking" when Human Understanding can only be limited to a view
2. The role of Universal Thinking in the Human Civilization



Post 1 by Siamanto

Personally, I think that we can think and communicate about what exists only using models - that philosophers have called ontologies and IT specialists call Data Models - where basic concepts - and their relationships - are defined and all "facts" are recorded and communicated using those basic concepts or others derived from the basic ones.
Similar concepts may be defined in a completely or slightly different manner in different models. Some models can more accurately describe/represent certain facts than others and models change, evolve and replaced when, eventually, they become obsolete; in other words, semantics are relative and there are no "Universal or Absolute Models."

Post 2 by axel

Some models can more accurately describe/represent certain facts than others and models change, evolve and replaced when, eventually, they become obsolete; in other words, semantics are relative and there are no "Universal or Absolute Models."

Just out of curiosity: are you using OMG's MOF as your meta-metamodel?

Post 3 by Siamanto

It is about M1-models or M0-models.
Just curious, what do you exactly mean? If you mean whether they are semantically compatible with MOF's meta-metamodel - i.e. can be described using an M2-model based on MOF - then what is the point or merit of the question?

In any case, for the record, the closest to what I have called "Universals Models" would be standards recommended by Oasis, standards such as JusticeXML or what L. Silverstone "pretends" to be Universal Data Models.
(I used the word "pretend" because they are not industry standards and there exists no consensus about them; it was not about the quality of his Data Models.)

Post 4 by axel

...
Another reason for mentioning MOF is that the more you go up the meta hierarchy the less the (m*-)model changes so I was hoping to find some stability by going all the way up to M3. (I for one, do not like MOF and its subsets or pseudo-subsets (EMF) very much. Way too limited. Way too static)

As to universal modeling, could we say it is a myth, a mere illusion, a mirage (speaking of oasis...)? Take the example of UML for instance. A complete failure. Even the complexity of software systems cannot be captured by "universal" metamodels. We are confronted with the eternal return of domain-specific stuff.

Post 5 by Siamanto

Another reason for mentioning MOF is that the more you go up the meta hierarchy the less the (m*-)model changes so I was hoping to find some stability by going all the way up to M3.

That's an abstraction into a meta-level. As I have said above, my goal is neither of epistemological nor methodological nature.
If it is of interest, please feel free to create a separate thread to discuss it.

As to universal modeling, could we say it is a myth, a mere illusion, a mirage (speaking of oasis...)?

It may be an illusion but it helps bringing different views closer and become more easily exchangeable, interchangeable and understandable; speaking the "same" - relatively speaking - language has practical consequences and it helps. However, believing in the absoluteness of a "Universal Model" is of religious nature.

You may have noticed that in my comment about Silverstone, "universal" was related to "enjoying consensus." (Are you familiar with Quines' concept of "inter-subjectivity" and how it compares - or equates??? - to "objectivity?")

Despite all, "universal modeling" is not a pure "myth, a mere illusion, a mirage" because it has shaped our civilization in the past and it continues to do so: it gives a direction.

(I for one, do not like MOF and its subsets or pseudo-subsets (EMF) very much. Way too limited. Way too static)

EMF has a completely different purpose than MOF and is not a subset of the latter - but, yes, as of today, is limited to a subset of it.

EMF is a framework offered within Eclipse to facilitate and support MDA based software development while MOF is a facility to semantically specify/describe modeling languages, facilitate and allow their interchangeability etc.
In other words, EMF is development tool while MOF is methodological tool and have different purposes and Use Cases.

I believe that EMF is - as of today - limited to a subset of UML for practical and historical reasons:
1. If you consider the features of existing MDA tools, you would notice that they offer limited capabilities and are mainly focused on Class Diagrams - some started to consider Activity Diagrams as BPEL is becoming more widely accepted
2. If you consider MDA driven software development in application shops, you will notice that it is limited to - in almost all cases - to the generation of simple Java/XML/DDL/C/C#/... artifacts from Class Diagrams - and in few cases, BPEL from Activity Diagrams

So considering
1. The reality of the industry and its immediate requirements
2. The fact that building a comprehensive framework that covers all of UML is an enormous project that
2.1 Requires many resources - that may not be unavailable in the Open Source community, or can be better used on more urgent projects
2.2 Has a high risk of failure - because of the size of the project
Initially, limiting EMF to a small subset of UML seems as a wise project management decision. Incremental development better manages risk. What do you think?

What do you mean by "static?" Thanks!

Take the example of UML for instance. A complete failure.

Considering how and why Software Development has evolved from "hacking" to become an engineering discipline - and becoming a science???, then I would not say that UML is a total failure.

It is beyond the scope of this thread to discuss how UML, Object Oriented views, Patterns, etc. have impacted the software development process; however, I would succinctly mention that UML - that is widely accepted in IT shops - helped improve:
1. Documentation and communication
2. Clarity of concepts
3. Quality of software
4. Reusability of software artifacts
5. Repeatability of software development
6....

Even the complexity of software systems cannot be captured by "universal" metamodels. We are confronted with the eternal return of domain-specific stuff.

Is your concern "universal metamodels" or (M1-layer) "Universal Models" as the ones that I have mentioned in my previous post? I would answer differently.

Also, what "domains" do you have in mind? Thanks!

Post 6 by axel

I was browsing the forum's main page when I saw your username appear next to the subject "does age matter?". I figured, he must be discussing Nietzsche and Data modeling. Let's have a closer look.

EMF has a completely different purpose than MOF and is not a subset of the latter - but, yes, as of today, is limited to a subset of it.

When I mentioned EMF, I was refering to Ecore (the bootstrapped M3 model) which more or less corresponds to EMOF (Essential MOF), a subset of MOF which was extracted from MOF in version 2.0, somewhat aligning the latter with the EMF implementation.

'Domain-specific':
When you attempt to capture some information, either you reuse an existing metamodel such as you UML and then profile it so as to make it fit your requirements* or you define a brand new metamodel which espouses the domain you want to target.
You can consider UML as domain-specific and it can be of help in certain contexts but clearly its pretentions were far greater at its inception.

* The UML profile even in version 2.0 is too weak in that regard. Providing such a mechanism is implicit recognition of the predictable (one-size-fits-all never works) failure of UML as a universal language. (note that the terms 'language' and 'metamodel' are often used interchangeably)

EMF:
EMF is not exactly a small subset of UML (or only to the extent MOF is). It is better not to consider it as such in order to avoid confusion (even though it bears a lot of similarity with the class diagram) if only for the fact it stands at a different M-level. EMF is actually used to define M2 models (aka metamodels) such as UML (an implementation of UML2 based on Ecore is in fact provided as a related project)

'Static':
By "static" I mean that the structural features of instances of M3-level types (often refered to as metaclasses) are statically determined by the M3 type being instantiated (=> all metaclasses have the exact same structure), you don't have a stereotyping mechanism (as in UML) let alone OCL-like boolean expressions that would allow for further specialization on an instance basis. (True with EMF you have annotations which roughly amount to stereotypes and associated tagged values but these have no underlying schema. Both values and (implicit) structure are specified on an instance basis)

Despite all, "universal modeling" is not a pure "myth, a mere illusion, a mirage" because it has shaped our civilization in the past and it continues to do so: it gives a direction.

Quoting from Gustave Le Bon (I gave the reference in another thread)

"Depuis l'aurore des civilisations les foules ont toujours subi l'influence des illusions. C'est aux créateurs d'illusions qu'elles ont élevé le plus de temples, de statues et d'autels. Illusions religieuses jadis, illusions philosophiques et sociales aujourd'hui, on retrouve toujours ces formidables souveraines à la tête de toutes les civilisations qui ont successivement fleuri sur notre planète."

[Siamanto]

Re: The Role of Universal Thinking In The Human Civilization.

I was browsing the forum's main page when I saw your username appear next to the subject "does age matter?". I figured, he must be discussing Nietzsche and Data modeling. Let's have a closer look.

Besides metamodeling, do you have any interest in M1-layer models that raise the issue of "universality" such as the ones that I have mentioned earlier - i.e. standards recommend by Oasis, JusticeXml...?

EMF has a completely different purpose than MOF and is not a subset of the latter - but, yes, as of today, is limited to a subset of it.

When I mentioned EMF, I was refering to Ecore (the bootstrapped M3 model) which more or less corresponds to EMOF (Essential MOF), a subset of MOF which was extracted from MOF in version 2.0, somewhat aligning the latter with the EMF implementation.
......
EMF:
EMF is not exactly a small subset of UML (or only to the extent MOF is). It is better not to consider it as such in order to avoid confusion (even though it bears a lot of similarity with the class diagram) if only for the fact it stands at a different M-level. EMF is actually used to define M2 models (aka metamodels) such as UML (an implementation of UML2 based on Ecore is in fact provided as a related project)

Thanks, for the clarification; however, if your focus is meta-modeling and not MDA driven software development then why can't we completely ignore EMF and limit ourselves to MOF/EMOF? Honestly, it can only enhance clarity; in fact, even if eCore is aligned with EMOF, do you see any advantage talking about EMF/eCore instead of MOF/EMOF?
Personally, I have far more interest in MDA than MOF; however, your focus, here, seems to be MOF.

'Domain-specific':
When you attempt to capture some information, either you reuse an existing metamodel such as you UML and then profile it so as to make it fit your requirements* or you define a brand new metamodel which espouses the domain you want to target.

One advantage of "defining a brand new metamodel which espouses the domain you want to target" would be full compliance to MOF; while a UML profile using UML extensions is, by definition, not not fully derived from MOF.

However, advantages of a UML profile would be ease of documentation, (human) communication and exchangeability/programmatic processing of metadata - i.e. tools that serialize M1-models based on the new metamodel may not be readily available etc. etc.

Would you consider full compliance to MOF preferable to the above listed advantages?

You can consider UML as domain-specific and it can be of help in certain contexts but clearly its pretentions were far greater at its inception.

One acceptation of "universal" would be "standard," "consensus based" and "readily interchangeable," but you seem to prefer the more "pure" understanding of the term??? LOL I'm an eclectic, "absolute relativist???"

* The UML profile even in version 2.0 is too weak in that regard. Providing such a mechanism is implicit recognition of the predictable (one-size-fits-all never works) failure of UML as a universal language. (note that the terms 'language' and 'metamodel' are often used interchangeably)

Yes, providing such a mechanism may be considered as a recognition that UML - as derived from MOF - may not fit all needs - for objective reasons - or "cultures" - for subjective reasons; even when, as you have suggested above, a metamodel, equivalent to the UML profile, may be built on MOF.

Again, as suggested above, is full compliance to MOF - as opposed to practical "conveniences" and advantages - the ultimate criterion?

Would you be satisfied if MOF included a mechanism to build metamodels from MOF as a replacement/alternative for UML profiles? If such mechanisms are offered and UML tools integrate them then the above mentioned practical "conveniences" and advantages would become legacy and of lesser interest.

'Static':
By "static" I mean that the structural features of instances of M3-level types (often refered to as metaclasses) are statically determined by the M3 type being instantiated (=> all metaclasses have the exact same structure), you don't have a stereotyping mechanism (as in UML) let alone OCL-like boolean expressions that would allow for further specialization on an instance basis. (True with EMF you have annotations which roughly amount to stereotypes and associated tagged values but these have no underlying schema. Both values and (implicit) structure are specified on an instance basis)

If I understand you correctly, it seems to me that you would like to introduce what you have considered as an "implicit recognition of the predictable (one-size-fits-all never works) failure" in the M2 layer - or maybe above????
Maybe, you have in mind something like what I have suggested above - i.e. a standard mechanism, as part of MOF, to derive metamodels that would replace UML profiles?

Despite all, "universal modeling" is not a pure "myth, a mere illusion, a mirage" because it has shaped our civilization in the past and it continues to do so: it gives a direction.

Quoting from Gustave Le Bon (I gave the reference in another thread)

"Depuis l'aurore des civilisations les foules ont toujours subi l'influence des illusions. C'est aux créateurs d'illusions qu'elles ont élevé le plus de temples, de statues et d'autels. Illusions religieuses jadis, illusions philosophiques et sociales aujourd'hui, on retrouve toujours ces formidables souveraines à la tête de toutes les civilisations qui ont successivement fleuri sur notre planète."

[A quick translation:
"Since the dawn of civilization, crowds have undergone the influence of illusions. It is to the creators of illusions that they have erected more temples, statues and altars. Religious illusions in the past, philosophical and social ones today; as sovereigns, they have headed all the successive civilizations that have flourished on our planet."]

Yes, that's what I meant. I would simply add that, in my view, "illusions" are often created by "poets" and "dreamers" while the statues are often erected in honor of those who use and abuse - them - the illusions - to shape reality. Of course, the same person can assume both roles.
In my late teens, I used to say: "Que dire de nos sociétés où l'ex-centrique est un/le centre de gravité?" Of course, the "ex-centric" being the poet, the "creator of illusions."
[Transaltion: "What can be told of societies where the ex-centric is a/the center of gravity?"]

How about: "The World was created in God's image, but God was created in the poet's image." Isn't God the Ulimate Universal Entity?

[axel]

Re: The Role of Universal Thinking In The Human Civilization.

A quick answer "en vrac"
The title of this thread sounds a bit pretentious to me btw
"The Human Civilization". There are a number of civilizations which have emerged, risen and fallen in history, each with their own distinctive cultural substrate. There is no such thing as THE human civilization.
Oasis (ebXML...), all these are metamodels / languages (M2). Instances of these metamodels, eg an ebXML document, are M1 models. I don't have any interest in such languages which are just as boring as XML schemas
MDA is essentially a marketing buzzword for an idea which has been around for ages.
Serialization of M1 models as XMI is free with any "MOF"-compliant metamodel (see EMF implementation)
UML is fully MOF-compliant.
I don't like MOF very much, as I said previously, but, in some sense, it is more universal than UML, as most metamodels can be expressed as MOF models.
Defining a metamodel is defining a language and agreeing on a common language in a given field is required for interoperability. that is kind of obvious.
A MOF model is a metamodel, on the other hand, a UML profile is just a hack to extend a given metamodel (UML) with additional (second-class) metadata.
Having everything based on a given meta-metamodel (MOF for instance) simplifies things greatly as all models can be manipulated in a uniform fashion with generic tooling

Some reading for you (I suggest you google for MDSD, DSL, DSM...)

http://www.voelter.de/mdsd-book/data//MDSDStahlVoelter-Chapter1.pdf

http://thebootstrap.blogspot.com/2004/06/stuart-kent-on-elaborationist-mda.html

http://acceleo.blogspot.com/2007/05/mda-vs-mdsd.html

How about: "The World was created in God's image, but God was created in the poet's image."

Well, this is an unorthodox (heretical) statement.

Isn't God the Ulimate Universal Entity?

God is beyond any definition.