FW: What I would like GME to be

[GME]

----------
From: 	Akos Ledeczi
Sent: 	Sunday, September 29, 2002 3:27:32 PM
To: 	GME Group
Subject: 	RE: What I would like GME to be
Auto forwarded by a Rule



[Adi]  

I will try and describe what I meant. I guess GME fits the Model/ View/ Control architecture with model -> mga file, view -> GME and controller -> com layer and I stand corrected. I was trying to say that as a programmer I can neither subscribe to GME events nor send commands or messages to GME. For example,  the search plug-in made by Sachin, displays the results of a search in a dialog box, but it cannot tell GME it bring up  the model when a user clicks on model in the dialog box. I have seen similar needs while using the decorator when I would like to catch mouse events and also perform some actions on GME. 

I summaries it I would like to say that GME should have an API access to control the viewing of the models.

[Akos Ledeczi]  

Oh, that's what you mean. A COM interface to the GME GUI has been on the TODO list for at least a year now. It is just a question of who will have time to do it... Peter: why don't you post the current TODO list, so people can see what's on it.


[Adi]   

I have some ideas as to how to realize  << File: Dream Meta GME.zip >> t. So I created a metamodel. It is a first cut at a new GME metamodeling philosophy. I would like to point out the advantages of having a UML based approach

1. The UML class diagram describes how the objects are to be organized physically. (Without having to bother about the visualization)
2. The Visualization specifies how the objects should be visualized in GME,  (The visualization is a separate aspect of the model). We can possible have different visualizations of the same models. 
3. This approach gives more power and flexibility, for example
	1. Ability to have a set which is also as a reference 
	2. Ability to have a reference to an association class and visualize it as an entity and many more such permutations
4. There Is a larger community that knows UML than the current MetaGME2000. 

I have attached a zip file that contains the proposed metamodel, an example model and icons. Please check it out and let me know what you guys think about it.
[Akos Ledeczi]  

This is one area where you will find great resistence from me ;-) 

- First of all, I think the current metamodeling environment is pretty good. I think it is easy to use, easy to learn yet pretty powerful. There are bigger holes in GME that could use improvement...

- Second, UML was invented/created to model object oriented software. GME metamodeling is different. So when you say that yours is closer to pure UML and that's good because a whole lot if people know UML, you must be talking about the syntax. Semantics, i.e. what a class diagram means, is a whole different game. In GME metamodeling context a class diagram has nothing to do what the same class diagram tells a software modeler. So, in other words, people would need to learn GME concepts and how they relate to UML class diagram concepts anyways. There is no way getting around it. Now, whether syntax-wise the paradigm is 99%  pure UML class diagram or only 80%, does not make a big difference IMHO.

- Third, coming up with a new environment that is better than the current one would take a whole lot of work and time. Similarly, the interpreter would be a major pain. Remember how many special cases came up in the first year of its use? Peter fixed a whole lot of bugs for months. What we have now is very stable.

- Now, the good news. In Janos' NSF project, there are a whole lot of issues related to metamodeling. He just mentioned it to me that we will need to support UML2, whatever that means. So, next year there will be a project that deals with metamodeling and it'll result in a new environment. (whether I like it or not ;-)

- We can certainly start a discussion now on it. Instead of everybody trying to figure out what you did from the attached zip file, it would be better to sit down with everybody interested and you could tell us your ideas. Next week looks pretty bad, but we could make it work Friday around lunchtime.

[Akos Ledeczi]