Integration & Web Services
the last couple of years, I spent a lot of time working with with EAI
packages like Tibco and Vitria to integrate applications across the enterprise.
The benefits are lucrative, but only if a large portion of the enterprise
can be integrated on a common information bus. One of the largest issues
is that each vendor has a proprietary solution that is being hailed as
that common bus. Web Services promise to provide a standards-based approach
to integration, but many of the features required for enterprise deployment
are not part of the standards and again reside in proprietary extensions.
In my opinion, the most interesting
part of Web services movement is the renewed focus on Service-Oriented
Architectures (SOAs). While the concept is intriguing, a lot of hype overshadows
many of the real issues to be resolved. My latest article
on SOA tries to shed some light on this.
I started to collect a set of enterprise integration
patterns. I found that many integration projects reuse a set of common
patterns, but very little of this knowledge is available outside of EAI
architect's heads. My paper on enterprise integration patterns was accepted
to the PLoP
2002 conference. This has now become the basis for my book Enterprise
Application / Integration
There is a huge amount of interest in .Net. I have
seen a lot of die-hard J2EE developers convert. I need to spend a little
more time on .Net.
is interesting to observe the application / integration convergence. .Net
is great example, Web Services-based integration is completely integrated
into the development environment.
last, we are seeing some better IDEs in the market. IntelliJ
IDEA is the best Java IDE I have seen. It has many built-in refactorings.
Eclipse is interesting
as well as people start to add integration-related panels (e.g. visual
If you haven't put a flame mail together yet for my
views on the first few topics, maybe you are ready to hit the 'Send' button
now. I firmly believe that an architect that has not sat down in front
of an IDE for a long time is losing a significant portion of his or her
effectiveness. I believe that a good architecture takes characteristics
of the tools and languages into account. That's hard to do if you have
never written a line of code. I do not deny that there are some very good
architects who never code, but I have seen a lot more bad ones...
I am convinced that one of the key factors to
successful software projects is the definition of a process that takes
factors like technology, tools, staff experience into account. Too many
people seem to rely on high-level, top-down 'methodologies' like RUP etc.
Nothing bad about the RUP, I like many of their ideas, but it does not
(and cannot) tell the project manager what to do next. Understanding the
technology in detail, knowing the tools and their quirks (I guarantee
there are some), understanding the solution architecture is the key to
my task planning. My task plan resembles the architecture. Systems architecture
is driven by dependencies and layers of abstraction. The same philosophy
applies to the task plan. I believe that a good architecture is a key
factor to a successful division of work, and a successful integration.
Currently, I am working with Martin Fowler on developing an agile approach
to EAI projects (see our LOMA talk).
My previous EAI 'Methodology' was built on top of ideas extracted from
RUP as well as a series of books written by Gerald Weinberg.
I spend a fair amount of my time creating presentations
and drawing diagrams. While I am often dismissed as a "PowerPoint
Jockey" or "Marchitect", I believe that the visual representation
of architectures or dependencies is one of the most neglected areas in
the field of software design and architecture. Using UML does not necessarily
guarantee readability. Creating correct, meaningful and appealing diagrams
remains a very difficult task. Maybe it is the combination of left and
right-brain interaction that makes it so challenging.
Human Aspects of
My interest in this area really started when we read
'Peopleware' by DeMarco and Lister for the Technology Management class
at Stanford (thanks to Prof. Tom Kosnik!). I agreed with almost every
sentence in this well-written book. It is hard to believe how many project
managers run a software project the same way they would run a meatpacking
plant. Software development is still a maturing discipline compared to
other engineering disciplines and human factors like developer experience
and the working environment are the biggest determining factors to project
productivity. Day in and day out I work in environments where people work
in noisy cubicle environments where their days are interrupted by 'status
meetings'. It seems like consultants end up at the bottom of the food
chain when it comes to work space -- surprising when you consider that
the client pays some $300 per hour for their services.