Speedo, JOnAS and ObjectWeb

[Roland Hedayat <roland@xxxxxxxxxx>]

Speedo, JOnAS and ObjectWeb
===========================

Two weeks ago, the following text (attached below) was sent
to the community list. It was later forwarded to the
architecture and board of directors lists.

The letter was written as a result of concern for the OW
strategy with respect to O/R mapping in general, and the
position of Speedo in the OW suite in particular.  Our
letter did not result in any answer from OW architects or
strategists and this has increased our concern.

For any enterprise, small or big, needless to say that the
enterprise data is mission critical. While JOnAS and the
other OW middleware collect, transform, distill, transmit and
present data, the O/R component is the glue between the
application server and the database, and thus an extremely
important component. A middleware offering including an
application server is not a complete one without a good O/R
mapping component.

Fortunately, there exists one with a very great potential,
namely Speedo, in the OW offering.

Unfortunately, it is our impression, as we pointed out in
our original letter, that this component is not visibly
attributed the importance and priority it deserves.

When we chose JOnAS/Speedo as our solution for the health
care projects in Stockholm City Council three years ago, we
took it for granted that Speedo, then in its infancy was to
be considered as crucial a component as the other central OW
components such as JOnAS.

We therefore ask the OW management as well as the board of
architects the following questions:

1. Is a persistence layer with an O/R mapping solution
considered an important component on the server side?

2. If yes, do you intend to keep control of this component,
both for technical reasons such as for example optimization
for usage with clustered JOnAS but also due to the
importance of having a complete offering to potential
users/customers?.

3. If yes, is Speedo the OW component of choice for
O/R mapping? Do we have a commitment from OW that this
component is going to be developed, maintained and
promoted by OW?

4. Or is there any other OW component with equivalent
functionality and performance that you recommend us to use
instead? Or do you recommend us to change to Hibernate?

If the latter is the case, we will have greatly increased
difficulties in explaining to our managers why not use a Red
Hat/JBoss/Hibernate solution altogether.

We have experienced problems, as described below. It is
our feeling, though, that this is mainly due to giving
too low priority to Speedo and its integration and
release synchronization with JOnAS, not to any serious
technical problems.

Being committed to using and promoting OW solutions, we urge
the architects and board of directors not to throw away a
true techo pearl "made by OW", and instead clearly state the
importance of Speedo within the OW infrastructure.

Best regards,

Mikael Karlsson
Creado Systems

Hans Höök
Höök Utveckling AB

Roland Hedayat
Inherit S AB

=========================================================

Our original letter from May 29th enclosed:


Speedo and ObjectWeb
====================

The signers of this letter are representing consultancies,
currently using Speedo in health care applications together with
Jonas as well as standalone. The areas where OW solutions have
been applied by us are both safety and performance critical.

The end customer for our solutions is primarily the Stockholm
City Council, organizing the public health care of approximately
1.8 M inhabitants.  Some of these solutions are even considered
for national health care projects in Sweden.

Three years ago, we were evaluating various open source
middleware solutions for use in a pilot system to be developed by
Stockholm City Council.  For us, we had decided up front to make
use of an Object/Relational mapping solution, and preferably a
standard compliant one.

Our candidates boiled down to JBoss/Hibernate and JOnAS/(Speedo or
Hibernate).

For several reasons, we opted for JOnAS/Speedo. One reason of
particular weight was the fact that Speedo at the time was part
of the OW suite (thus giving us a kind of "one stop" offering)
and also compliant with a standard (JDO), while Hibernate was an
independent 3rd party product, not compliant with a standard.

So, the very existence of a good O/R mapping solution in the OW
suite of solutions was among the important factors in favour of
OW on our check list.

We chose JOnAS/Speedo, and we have not regretted it.

We came to be early users of Speedo, but the development team has
been extremely helpful both with technical advice, solving
problems, correcting bugs we found etc. The product is evolving
towards high reliability and performance. For us it is a key
component.

However, since the early days, things have happened:

* JDO did not become _the_ one and only standard for Java persistence,
  it will coexist with EJB3.

* Hibernate is now part of the JBoss offering

* Hibernate is in the process of being standardized according to EJB3

Does this make Speedo a less important or interesting product for
ObjectWeb?

The reason for asking this question is that we have got the
feeling, based on some indications, that Speedo is not considered
a central, core part of the OW offering:

  - When a new release of JOnAS is out, it is seldom synchronized
    with the development of Speedo. In many cases one even has to
    build the latest Speedo distro and integrate it manually into
    JOnAS

  - The Speedo development pace seems to have slowed down lately,
    mabybe because of lack of resources

  - There is little promotion of the Speedo component on the OW
    site

Our answer to the above rethorical question is NO, for the
following reasons:

1. A good O/R mapping solution is increasingly important to
   bridge the "semantic gap" between a good server side
   architecture, based on OO principles, and the still prevailing
   relational database.

2. If one has to go elsewhere to find this solution, then the
   probability is much higher than one will find another solution
   (than the OW-based one) altogether. Why, as a newcomer, choose
   JOnAS and third party JBoss Hibernate rather than "one stop
   shopping" a JBoss/Hibernate solution?

3. It is in many situations nice to have an O/R component which
   is not bound to the EJB stuff, that is, can be used stand
   alone where appropriate.

4. JDO2 is now a mature standard, and in some respects superior
   to the EJB3 standard for persistence. The JDO2 standard will
   probably continue to live for the foreseeable future, even if
   the long term goal is to merge the two standards into one.

5. Speedo is one of the few Java persistence frameworks which will
   support _both_ JDO2 and EJB3.

6. Speedo is in itself a showcase for the high quality
   architecture and engineering so often found in OW components:

   - Highly modularized by means of Fractal, reusing Perseus,
     Medor, Jorm and relying on the ASM tools

   - High performance

7. Maintaining an own O/R solution makes it easier to optimize it
   for usage within JOnAS

After having used OW solutions to our and our customers
satisfaction, we actively promote ObjectWeb. For us, Speedo is a
central part of this, and we hope that the ObjectWeb Consortium
will continue to support Speedo, and also clearly promote it as a
first order citizen in the suite of solutions. It so deserves.

Mikael Karlsson
Creado Systems

Hans Höök
Höök Utveckling AB

Roland Hedayat
Inherit S AB

-- 
==================================
| Roland Hedayat
| Inherit S AB
| Långsjöv. 8
| 131 33 Nacka, Sweden
| roland@xxxxxxxxxx 
| Tel: +46 (0)8  641 64 14 
| Mob: +46 (0)708 18 07 69 
| Fax:  +46 (0)8  718 52 88