Sunday, December 19, 2004

The Four Tenets of SOA

As we near this Christmas holiday, it seems only appropriate to throw the bible into my opponent's face :-)

Consider the following:
'You shall love your neighbor as yourself.' - (can, shall, may, etc.)
'Don't bang your neighbor's wife' - (don't, shall not, etc.)
Subtle difference? I don't think so. Imagine the first "rule" without the second "rule". Talk about a HUGE clarification!

You see, leaving room for interpretation leads to problems. That's why the Four Tenets of SOA are utterly useless (even as philosophies). Don't do X; Don't do Y. Don't do Z. Simple isn't it? Constraints. Ok, hopefully I've established the difference between philosophy and constraints. That said, let's go back and take a look at the four tenets that Don Box provides us peons to guide our professional existence for the next decade:

Boundaries are Explicit
Point: Because each cross-boundary communication is potentially costly, service-orientation is based on a model of explicit message passing rather than implicit method invocation.
Response: This is the "distribute OO was a failure" rule. Rather than saying
A. Distributed OO was a failure because Microsoft screwed the CORBA guys and fragmented the space.
B. Distributed OO was a failure because CORBA was a pain in the ass and it was competing in a rough market (BPR with 2-tier RAD and RAD groupware).
C. Distributed objects was a failure because we didn't adequately factor out the non-functional concerns and network programming was in its infancy.

No, rather than saying A,B or C - Microsoft says "Distributed objects failed because programmers were so stupid that they couldn't tell the difference between local and remote calls. Their solution was simple. Make remote calls a pain in the ass (oops, I mean ... make them non-primitives.) Yes, create some sort of language add-on (an API?) that mandates a mapping layer between the internal type system and the 'generic' type system - that'll teach those stupid programmers to realize when it's local and when it's remote! Oh, but let's also build runtime optimization of local service calls into the Indigo stack just to mess with them!!!
Sweet.

In case you're wondering, the constraint for 'boundaries are explicit' is this:
"language designers and extension builders SHALL NOT create a programming extension or primitive that enables service calls to be or appear to be first order primitives." (This might be the rule, but I can't wait to break it!!!)

The second part of 'boundaries are explicit' is: "The notion that boundaries are explicit applies not only to inter-service communication but also to inter-developer communication...." For the life of me, I can't find a rule, philosophy or anything actionable in this one.

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

Services are Autonomous
This is a case of bad terminology and description (I believe). If I understand Box correctly, his real intent was this line, "services are almost always deployed atomically". Whew... that was hard. Let's cut to the end and find the constraints:
1. Newly versioned services MUST BE deployed in a manner that DOES NOT negatively affect the prior version.

The second part of 'services are autonomous' states: "Service-orientation encourages a model that increases ubiquity by reducing the complexity of service interactions."
Here, Box takes jumps out of the versioning issue and lands in the 'loosely coupled' issue... although, in perhaps the most ambiguous manner possible. All I can offer is that if you want loose coupling to be a tenet, than you have to make an effort to create an index on loose coupling . Saying that 'leaking intentions is bad' doesn't quite do it for me.

==================================================================
Services share schema and contract, not class

Box states, "Rather, services interact based solely on schemas (for structures) and contracts (for behaviors)." Schema and contract only? Crap, we have to kill SOAP with Attachments and DIME ;-)

I've stated before... I don't have a good answer for this one... the problems with this tenet are clear: sharing binary information, passing predicates, bypassing the schemas with optimizations and sharing instructions other than classes like DSL's and metalogic.
==================================================================

Service compatibility is determined based on policy

Again, I agree.
Rule: Service consumer MUST be able to determine the capabilities of a provider via a publicly accessible, standard policy.
==================================================================

Now, I know that Don Box didn't mean for these to be rules... just some high level philosophies. However, if you watch the MS video, you'll see the MS-band-of-jokers butcher his material.

Back to my original point; we need simple rules or someone might accidentally misunderstand things like 'love your neighbor'. It's all about constraints. :-)

No comments: