Re: More basic questions please

Hi,
please see my comments below:

BTal wrote:

> I would like to know please if there is a road map for MDR. What are the
> plans for this tool in the foreseen future? Are you working on a new
> version?

In the near future we plan to do an architecture review of MDR, 
finalizing the set of APIs. As a result of that we may break the 
compatibility btw and I plan to clean a few things and add make the 
"planned change" events vetoable. This is required in order to make MDR 
an integral part of NetBeans releases. We are currently working on a 
metamodel of Java and its implementation. All the Java metadata (about 
classes in currently open projects in NetBeans) will be available via 
JMI in NB 4.0. This will also enable us to implement refactorings.

> Also will it upgrade to support MOF 2.0? Will it use CMOF (in opposed to
> EMF's usage of EMOF)? Will MDR support UML 2.0 conversions?

We are still not sure whether to upgrate to MOF 2.0. As MOF gets more 
popular and one of its requirements were to share the same core with UML 
2.0, too many people started to add new stuff, change everything from 
the roots, which made both UML and MOF very complicated not only for 
tool vendors, but also hard to understand for users. We are now trying 
to fix this in finalization task force in the OMG. If that does not 
succeed, we will probably implement EMOF only (if we will see that it is 
likely that it will get adopted by people). It will take several months 
to make the decision since FTF is not finalized yet and JMI for MOF 2.0 
has not even started.

> IBM claims at the EMF faq that: " The Eclipse EMF Java mappings have been
> optimized for in memory tool integration using Java and XML while JMI has
> been optimized for metadata repository usage scenarios". Is this the way you
> guys are looking at the comparison between the two (very similar) products?

I would agree with the first part of sentence - EMF seems to be heavily 
driven by what you can do in Java and XML. So it is more like a simple 
model of XML. And it may proof itself (and maybe it is already the case) 
that it is enough to do an efficient metamodeling. MOF was more driven 
by what we need in order to have a good meta-meta-model that could 
easily by mapped not only to Java, but also to IDL and other languages 
and what construct it would be nice to have to navigate the metadata 
efficiently. It explicitly defines lifecycle of objects, identity, etc. 
so that these things are the same no matter which platform your 
metamodel is mapped to.

> Reiterating the question in the thread below, do you have plans to have MDR
> evolve into a MOF central repository system/tool? I noticed the JDBC work
> being done lately, but I was wondering about the importance of this aspect
> in your road-map.

For NetBeans the JDBC storage is not that important (if it was, we would 
put resources into developing it ourselves). Our main usage of MDR is 
very similar to IBM's usage of EMF - we want to use it as a (meta)data 
integration platform in our tools. So the fast, not necessarily too 
robust and tightely integrated b-tree with an efficient paging mechanism 
(to avoid high memory consumption) is more appropriate for our usecase. 
Currently we are redesigning the Java module to use MDR to expose 
metadata about Java applications open in NetBeans to other modules. 
Other modules will hopefuly follow.
Regards,
Martin