SOA meets the real world
New in
InfoWorld -- "
Five big organizations launch ambitious service-oriented architecture initiatives and explain their objectives, obstacles,
and solutions."
What I like best is at the beginning:
SOA turns the conventional model of enterprise
software development on its head. Normally, programmers write software
based on a set of well-defined requirements. SOA demands that
organizations create an ecosystem of services that may ultimately have
an army of stakeholders inside and outside the firewall.
All the technology debates that circle around services orientation
--Java vs. .NET, REST vs WS, simplicity vs complexity -- are beside the
point if people don't grasp the basic point that the really hard thing
to do is come to that vision of an ecosystem of services. That's
where the real benefits -- and the big costs and risks -- lie,
not in the plumbing.
A couple of random thoughts after reading through them:
- "Protocol independence" is scorned by REST advocates, but notice
the various organizations that had to deal with the reality that MQ and
JMS transport systems were pervasive and nobody apparently wanted to
replace them with HTTP. In fact, I can't find the string "HTTP" in any
of those articles.
- One of the tenets of REST is to identify resources with URIs;
note the hospital case where the really hard part was to figure out
which unique person the various bits of data referred to, not whether
to identify the people with URIs or database keys or whatever.
- None of these systems ended up being all that loosely coupled as far as I could see.
- None of these were "on the Web", so they just support the idea
that web services approaches have success stories in the more
controlled world of enterprises with management hierarchies and legacy
systems that work, even if they don't interoperate.
There's still no evidence that these approaches would scale to the Web, or
would be appropriate for new Web-based applications of the sort
that MSN, Yahoo, Google, and a lot of smaller companies are
building.
[Quick correction added: "NO evidence that these approaches would
scale to the Web." Leaving out the "no" changes the meaning
rather drastically!]
