IMS Barcelona: SOA session

I got back from IMS Barcelona today, and wanted to post some reflections on some of the things said in the SOA session. I would have live blogged this, but the internet connectivity at the conference wasn’t great (which must have been stated ad-nauseum on twitter – or would have been if people had internet access to get to twitter…).

The speeches gave a good potted view on what the hype of SOA was back in the day, and followed up with a dose of what happened next. The premise was largely that SOA was dead, long live SOA. Chuck Severance gave one of the best service view presentations I’ve seen, and he was concluding that we have service orientation now – via REST apps that appear anywhere on the net at will. He also baited the audience by declaring that SOAP largely gets in the way of service oriented approaches, and rather everything should be REST…  Although on a previous point he’d said that service orientation was largely technology neutral. I think I agree with him there, but i’m a bit too old school to say that everything should be done in REST. I think  that in a year or two REST will have been replaced by something else – but the notion of a service isn’t new, so I think the service oriented view is one that will stay around. The consensus then was that SOA/soa isn’t a technology problem.

A large chunk of the discussion was regarding SOA in Education, and there was discussion on why it hadn’t really had that much impact. I don’t think any earth shattering points were made:

• “Not invented here” exists
• HEIs are just too complicated to be able to model
• The people writing the services have ignored the requirements of the HEI
• SOA means a load of cash upfront

I thought that point two – “HEIs are too complicated”, was particularly interesting given the work that we’re doing in the CET, and the work that is going on in the UK with Business Process modelling and Enterprise Architecture. Colin Smythe said right at the end that IMS were looking to do some of this BPM work with HEIs. Colin also made the point that it took ten years to model an aircraft carrier (and it was implied that said aircraft carrier was the same or less complexity to an HEI… or perhaps that all aircraft carriers are the same… unlike HEIs…)

I empathised with the point about organisations wanting to ‘buy’ an SOA, and thus purchase service development from big integrators, only to find that the service that they have bought does 80% of their requirements (if they’re lucky). This was further compounded by the integrator informing the buyer that, “this is what the service does. What you are asking for is not part of the service”. A particularly unhelpful attitude which almost certainly hasn’t done the service based approach any favours.

The final point, that upper case SOA is a big investment upfront, versus a more agile little case soa, was interesting. In essence, in an enterprise service system, the contract is so important that it has to be negotiated upfront, set in stone and then rarely or ever changed. The agile approach – pay as you go – means that all one needs is a protocol and many APIs – and you keep developing more and more stuff until you achieve your goal or the money runs out. Personally, I think that there needs to be a balance between how much is specified up front and how much is specified as you go – but then this is true of any software development.

I still think service based is the way to go. I think the re-use argument is so powerful in these times of slashed budgets. I also think that perhaps the current from of the service oriented approach is a victim of the times – which I’d put down to big business, the government’s increasing need to solve the most complicated problems in a single breath, and the developer revolution.