Continuing Review of a REST Based Microkernel

Pete Roger's founder of 1060 research emails me after my last blog entry, he writes:

We've been in stealth for a long time (3 years at HP, 18 months whilst negotiating with HP and starting 1060), so it's great to finally be able to discover if we've got something useful.

That's precisely what I'm trying to discover here. See, most products are easy to evaluate since they usually belong to a category where you can find "comps". However, the 1060 NetKernel product is one of those entities that claim paradigm-shifting capabilities. It's imperative that you don't get blown away by those high-falutin words.

If I could summarize the innovation introduced by this product it would be this (taken from the documentation):

What if the REST interface (URI address space) didn't end at the edge of your external interface?

In fact, the documentation describes the entire product as "A URI request scheduler". The aspiration is that such a generic abstraction can lead to "applications [that] are modular, scalable, adaptive to change, and maintainable". The simplicity is extremely compelling, however for innovative architectures like this, it's vitally important to demonstrate a live application (i.e. the proof is in the pudding deal).

The good news is that there are two applications that have been built with this framework (i.e. Bugxter and Blogxter). The bad news is that Blogxter isn't available yet for evaluation and Blogxter is being refactored to work with the latest release.

Nevertheless, without gaining an appreciation of the effectiveness of the NetKernel conceptual model, I have a few quick observations.

• Declarative Processing Markup Language (i.e. DPML) is a language for building declarative XML applications. I've never liked XML-based languages, furthermore I expected but could not find concurrency constructs in the language.
• The orthogonality of transport with service and representation is refreshing to see, I've discussed this issue in an earlier blog. Now how difficult would it be to have the configuration accessible via a web interface?
• Administration appears to be implemented using the framework so I'm extremely curious as to how this worked out. The original JMX implementation was deprecated.

I hope to gain insight on this new RESTful paradigm with more indepth study of the existing applications. It's a big learning curve to appreciate how to build effective RESTful applications. The pay-off however may be greater than what one would expect, the simplicity of the approach could in fact lead to applications with better manageability.

P.S. Anyone know if Zope is RESTful? It sure feels that way!