AW: [enhydra] New Look&Feel of Enhydra Site

["Madl Alfred - Together AT" <a.madl@xxxxxxxxxxx>]

Hi Lofi !

At our last developers meeting we had a very intense discussion (after 
detailled analysis by our guys) about UML vs. DOML and ended up with roughly 
the following results:

UML currently does not have enough semantic elements for completely 
describing a physical relational data structure with all its aspects like 
keys, (clustered/unique) indizes, and all the necessary additional 
information for mapping these structures to generated Java and SQL code.

Of course we could start using freely definable UML values (just single 
valued are possible !) and some numbers of rather ugly naming conventions to 
somehow "force" the information into UML and get it back again in the 
generator, but than again we would have no tool in the world to support that. 
Users would than have to learn how to "DOML" in UML...no good idea...no 
standard at all anymore.

But of course we can do UML to partial_DOML and (as we do already today !) 
DOML to UML/XMI transformations.

We also investigated lots of things in the area of CWMI, MOF, Eclipse 
projects for modelling, etc. but had no success to find something appropriate 
to build upon.

I think the basic issue is the principal approach: 

DODS is mapping "relations to objects" (and not "objects to relations" like 
all the other tools)...starting from a (hopefully) perfect and performant 
physical database design and simply mapping every table to a set of Java 
classes. 

In our view this is the only task a mapping generator/runtime has to fulfill 
and starting from bottom upwards in this area (at some point in time of a 
project and of course after lots of basic analysis work...) is the only 
chance to get results that perform and scale well. I could not think of 
another approach for that for e.g. our workflow engine "Shark" (which uses 
DODS) to handle more than 400 transactions per second and more than 2 million 
process instances per day in an installation we currently have...

We now started to clean up the old DOML schema and to extend it to fully 
support different key models (oid's per DB/DBGroup, oid's per table, IDENTITY 
columns, natural/multicolumn primary/foreign keys, etc.) to be able to use 
DODS also for legacy datamodels and to wrap these structures into proven and 
portable relational-object mapping code that really performs and scales.

In the end I think we will also have our own graphical DOML editor...but we 
have experience now in this area when you look at our new JaWE 2.0 :-)

By the way: Wouldnt it be nice to merge the EJOSA homepage into 
www.enhydra.org ?

Greetings.

Alfred

-----Ursprüngliche Nachricht-----
Von: Lofi Dewanto [mailto:dewanto@xxxxxx]
Bereitgestellt: Sonntag, 08. Jänner 2006 13:21
Bereitgestellt in: Enhydra
Unterhaltung: [enhydra] New Look&Feel of Enhydra Site
Betreff: [enhydra] New Look&Feel of Enhydra Site


Hi All,

first of all: Happy new year to Enhydra Community! Hope
to see a lot of new things coming in Enhydra in 2006!

I really like the new look&feel of Enhydra site, cool and
very easy to navigate!

I also read this part:
"A graphical DODS editor based on UML, possibly replace DOML schema with
UML/XMI"

This would be very nice! IMO it's worthwhile to base this on UML since:
1) You have some UML tools available (ArgoUML, Poseidon,
    MagicDraw, ...).
2) You need not to build your own tool for building DODS application,
    so you can save resources and use it somewhere else.

The next step would be to write an AndroMDA cartridge (MDA) to let all
the DODS file automatically created/generated. Please see the example of
Hibernate cartridge:
http://www.andromda.org/andromda-hibernate-cartridge/index.html

EJOSA (http://ejosa.sourceforge.net - now also with new look&feel :-))
has already an AndroMDA cartridge for EAF/XMLC presentation layer.
See: http://www.jaxmag.com/itr/online_artikel/psecom,id,653,nodeid,147.html

If we have a DODS cartridge, we will be able to create a complete
Enhydra Application (EAF/XMLC with DODS) completely based on UML. This
would IMO enrich Enhydra products.

At the moment we are updating the cartridges (also the EAF/XMLC) to
AndroMDA 3.1. All the examples (PiggyBank) will be also updated to
Enhydra Enterprise 6.5.x. I'll inform the community as soon as we 
finished this work.

Cheers,
-- 
---------------------------------------------------
Blasius Lofi Dewanto
---------------------------------------------------
OpenUSS - Open University Support System
http://openuss.sourceforge.net
---------------------------------------------------
E-Mail   : dewanto@xxxxxxxxxxxxxxx
---------------------------------------------------