SOA - Less coding, less outsourcing?

I always find Joe McKendrick's posts worth reading, and his post on 'Less coding, Less outsourcing' is no exception. Basically the points being made are that reducing coding will reduce the most commonly outsourced piece of the puzzle, and that better alignment of technologists with business-oriented end users will increase the likelihood of development being kept in-house.

I would like to add one further perspective to this discussion, however. A counter-balance to these two points may be that once a company has implemented SOA, it becomes much easier to look at a model based around Software as a Service (SaaS). There are already examples of SaaS success, such as salesforce.com, but one of the problems in this way of outsourcing a segment of business operations support is the difficulty of 'carving it out' from existing implementations. In order for SaaS to work, there need to be clear boundaries between the outsourced service and the rest.

Implementing an SOA provides this clean separation of discrete business services, linked together with standard interfaces. This immediately makes it easier to go down a SaaS road - the task is certainly not trivial, even in an SOA environment, but it is certainly a lot less effort and introduces considerably less business risk.

This may seem an odd position for me to take after suggesting recently that SaaS is not likely to be a killer app for SOA. However, I am not trying to say that SaaS alone will drive SOA adoption, but rather that once SOA has been adopted, SaaS becomes a more practical option.   

Steve