Wednesday, July 09, 2008

More duck with your pheasant?

One minute standards are too early, the next Bert tells us that standards are right for the cloud - here's theirs. Don't worry, Bert has given us something new to think about in his post on Open Source does NOT provide interoperability.

Of course, arguments just like portability need more than one participant. I'm not aware of anyone who would argue that Open Source, on its own, brings interoperability. What is sometimes discussed is how open sourced standards (i.e. complete open sourced operational reference models for a service) can be used (along with multiple service providers, trademarks and compliance authorities) to create competitive utility computing markets which have interoperability and portability between service providers.

Open source is key to this, it is critical for competitive markets and it is the only practical way of achieving it without entering a lock-in nightmare of a proprietary reference model. If Bert actually wanted to create an argument he could have said that Open Source is not key for interoperability in the cloud, then we'd have something to discuss.

I'm somewhat bored by all the vendors who try to explain why those components of IT which are ubiquitous aren't going to end up as services and who attempt to bring product mentality into a service world. The argument runs counter to an economy based on services. It's like those half baked ideas which pronounced the death of IT through commoditisation and ignore the feedback created by standard components (from commoditisation) which in turn leads to even faster innovation (as per the whole theory of componentisation).

However, since Bert is picking on open source, I thought I'd point out a few probable flaws with his argument:-

"Although there are open source systems for the base technology, they aren’t a complete solution". Well it's no surprise that without open sourced standards there is no interoperability. The closest we've seen to open sourced standards are the open SDK of GAE (Google App Engine) and Eucalyptus.

"Each provider must complete the solution themselves". Fortunately an open sourced standard is a reference model, it provides all the primitives. You can either implement it, or something that matches it. For example, if you consider that the open SDK of GAE is an open sourced standard, then GAE is simply Google's implementation of the open SDK and AppDrop is another.

"providers are looking for competitive advantage". There are two forms of competitive advantage here. One is in operational advantage, as in a faster, cheaper or somehow better implementation of the standard. The second is in product differentiation (a new feature etc). Product differentiation can* kill interoperability and it belongs in the product world and not in the world of ubiquitous activities treated as services.

"debian to ubuntu, Suse to Fedora, Redhat to CentOS": these are all competing products in a product world. In a service world you could get marketplaces such as "Fedora" as a Service where you will get Fedora and not something else. You could swap between service providers and know that as long as they complied with the "Fedora" standard it will work. Interoperability can be achieved with a proprietary standard, but it means the service provider gives up some strategic control of their business to the technology vendor. The use of open source in the standard is also about protecting the service provider (such as small ISPs) and reducing their barriers to adopting a standard.

I have no financial or commercial interests in the cloud world, I do however see lots of vendors either trying to persuade business consumers that interoperability, lock-in and portability are not issues, or that some committee will save the day.

I do wish Bert the best of luck with his standards committee, however if you are a consumer of these services, you really need to ask yourself whether you want the poachers to be the gamekeepers in a service world or whether it's business consumers who should be running any standards committee.

*I've added this conditionality in response to a comment by James Urquhart, with which I completely agree.