Stefan Tilkov pushes back on my assertion that REST significantly reduces the amount of custom coding you need to do to get a distributed application conversation going. On reflection, I think he is right. Assuming a rich set of tools for both paradigms (object/RPC and REST), setup time should be much-of-a-muchness between the two.

Where REST shines is in the other 90% of the conversation lifecycle when, in order to keep talking, you need to change things. The stability of the universal exposed application interface (GET,POST,PUT,DELETE) stands in stark contrast to the unstable, domino-effect-laden per-object interfaces endemic in RPC approaches.

The real cost of e-business integration lies in the 90% of the lifecycle that kicks off after the system is first up and running. Typically that 90% is termed "maintenace" but it is nothing of the sort. Maintenance is changing the oil in your car. Software doesn't need oil changes. It needs to be changed as business needs change.

With REST (and SOA) what we should be doing is optimizing our software development paradigms for the latter 90% of the lifecycle not the first 10%.

posted by Sean 12:08 AM

[Link] . . .

The slides from my talk last week at the XML UK event in Cambridge are here [850k PDF]. Powerpoint (saved out out the OpenOffice format here. The original Open Office file is here.

posted by Sean 12:23 PM

[Link] . . .

Close integrations of the third kind is an ITWorld article that looks at the idea of protocol negotiation - common in the lower parts of of an EAI stack - and their applicability higher up the stack, closer to the business. Think "business process negotation" rather than "protocol negotiation".

posted by Sean 12:47 AM

[Link] . . .

Simon St. Laurent posts some hot beefy goodness on the subject of CMS.

posted by Sean 9:42 AM

[Link] . . .

No April Fool's joke this.

In an attempt to explain my approach to web services and service oriented architectures, I have resorted to that refuge of the scoundrel known as the novel. A novel written in sections and published as I write them in blog form.

Herewith the first installment of (giggle now) A SOAP Opera.

I aim to write a new installment every week. I doubt if I'll be able to keep to that timescale but at least its a plan, right?

posted by Sean 12:17 PM

[Link] . . .

No modelling technology in history has, as far as I know, illustrated how the Sapir Whorf thesis applies to geek-speak better than XML. Some time ago, I wrote about modelling bishops in XML - illustrating how the language you speak influences how you create data models.
Last Thursday I received my advance copy of the travel guide to Molvania. I'm trying to learn enough of the language to approach a data model for a cheese manufacturing plant.

posted by Sean 9:12 AM

[Link] . . .

Ah! Nice. Now we have a meta-model for message choreography. Seems to be fully Omega Point Theory compliant. Wonderful.

What a pity we have *yet* to agree on a way to send a message reliably from one place to another though:-(

I think its high time we instaniated an infoset in which an element whose local local type is named "cart" occurs after an element whose local type is named "horse".

Wouldn't you agree?

posted by Sean 8:49 AM

[Link] . . .