One of the things that have been baffling me for the past few months is the amount of money and effort the big middleware vendors - IBM, BEA, Tibco, Oracle - have been pouring into their SOA governance products in terms of development and marketing.
It's not that I think that SOA governance is not important. Of course it is. But it is more of an organization, planning and consistency issue for those implementing SOA. The products themselves -- registries and repositories, etc. -- seem to be low-value products that are rapidly being commoditized and undifferentiated.
Seems to me that the more interesting issue, which few vendors are tackling, is how to implement your services. Existing approaches using J2EE and databases won't cut it in an SOA world where services need to be loosely-coupled, dynamically scale and so on. I previously discussed some of the challenges of implementing services in this post.
The good news is that from discussions with folks at some of the large vendors, it seems that some (senior) people are starting to get it.
By the way, Nati gave a great presentation, together with Shay Banon (one of our lead architects) on the topic of implementing services at TheServerSide Java Symposium last week, entitled From Tiers to Services without Web Services. It roughly follows the same concepts Nati discusses in The Scalability Revolution white paper we published recently.