Friday, September 02, 2011

The battle that wasn't ...

Chris Boos wrote an interesting post about a debate that @samj and I were having on twitter regarding APIs in the cloud space. I thought I'd leave my comment here as a general view on the subject.

A couple of things to point out. Twitter is not the best tool in the world to determine the exact context of a discussion because those listening aren't generally privy to the history of the discussion. Hence in this case, it may not be clear that Sam Johnston and I are in absolute agreement on the importance of open source and efforts like OpenStack in this world.

Any difference between us is on the necessity of reverse engineering APIs and co-opting as the main short term tactical play. The long term we're both totally in agreement on - open standards, open formats and open source are critical.

Our difference in views on short term tactical plays hardly constitutes a battle but is merely debate. As for being a "giant", whilst that is very flattering it doesn't coincide with my view of the world. Nevertheless, it was an excellent post by Chris and much appreciated.

Comment --

Just to clarify my view - as it currently stands any company can reverse engineer an API for reasons of interoperability. Hence when trying to make a market of providers in the IaaS space with semantic interoperability between providers, I strongly support adoption where there is clearly a dominant API.

It should be noted that such a market can have multiple open source and proprietary implementations around the API. However, running code through an open source effort is necessary to form a market place without a single (or consortium of) vendor(s) being able to force a tax on that market. In other words, providers need to have an operational means of implementing the service and compete in the market without a necessity to purchase software licenses (a tax on competition). They may choose to buy software to do so but a free market is one unencumbered by such forced taxation.

This is why I do no support MSFT Azure's effort, despite the provision of open standards because there exist no open source implementation.

This is why I did not support Google's AppEngine, despite the provision of an SDK as there existed no fully operational open source means of implementing the service.

This is why I strongly support open source efforts which reverse engineer the dominant API for reasons of interoperability e.g. open stack, eucalyptus etc.

It is also why I strongly support open source efforts which attempt to create the dominant standard in a fledgling market, such as CloudFoundry in the PaaS arena.

Once the marketplace of alternative providers is large enough and it has the dominant ecosystem then the open source effort in effect becomes the defacto standard for implementation and the API in that market. If necessary, due to abuse of position by the original provider, then the API can be differentiated away from the original provider including providing an entirely new API where applicable.

I don't find attempts to differentiate on API in a utility world where one API is clearly dominant meaningful. Of course if an open source effort (such as openstack) creates a large enough ecosystem then it is in effect the dominant and can do as it pleases.

I find re-inventing the wheel by creating an API by committee and attempting to get the market to adopt as a wasted effort when a market has in principle chosen.

I do find the way to standardise is through creating the largest ecosystem and in such cases both reverse engineering the dominant API for reasons of interoperability combined with provision of open source running code is necessary.

Co-opt rather than compete is the order of the day in this world.