Internal vs. external Systems, Part 2

"The design tensions for external facing systems, especially security, are very different from interal systems." I agree. But my point is that the tensions facing external systems are largely a superset of those facing internal systems. As a commentor said on the other post, "As a general statement you might want to say something like on average, architects are underappreciating the need for integration in the future".

I've built large SOA systems (for the Intranet), and I've built large RESTful systems (for both the Internet and Intranet), and there's no question in my mind which approach is a *lot* more conducive to unanticipated integration.

I agree with some of the differences you have highlighted but definately not the assumption that internal systems are somehow more secure than those exposed to the Internet. Yes its true that there are sometimes different threats to mitigate, but ask any security guy where they think the majority of explotations occur and I suggest that you will find that the answer is 'from within'. Also they're often harder to resolve because of the basic 'software fortress' assumptions that are made.

Anyway, back to the other point. I absolutely agree that only a subset of internal systems are likely to made available for external consumption. If we accept that, then going back a post or two, it picks up your theme of whether we should (in those cases) work hard for protocol independance or even technology independence. Large organisations often make huge investments in core technologies and have absolutely no intention of switching to an alternate vendor any time in the near term. They also want to leverage as much as possible out of that investment and if that means a degree of vendor lock-in and some loss of reusability, this is a price that may be worth paying versus a lowest common denominator approach, that has ultimate flexibility/replacability but suffers in other functional and non functional aspects. For years we have heard of the holy grail of a completely commoditised infrastructure, but the reality is very different and until the disruption and risk is minimised to the extent that moving from one supplier to another is very much smaller than it is today, I for one find it difficult in some cases to justify the extra design/build effort.