Prompted by the recent infoworld podcast, ManAmplified has posted a great blog entry that discusses his high-end enterprise deployments of NetKernel. He describes their NK hosted JMS integration platform as the 'jewel in our infrastructure'. Thanks for the kind words Chris.

Jon said we're 'a long way from mainstream'. Or...could it possibly be... that the mainstream is a long way from us?

Rui has developed oceanos an open-source 'blog' framework on top of NetKernel. The intial implementation is a blog application but the ambition is much broader - to create a distributed pluggable software framework that brings ideas from Eclipse over to a NetKernel hosted system. The first cut is already a fully functional blog. Good stuff Rui. Looking forward to hearing more at the NetKernel Architect's Weekend in September.

Which reminds me. I set myself a task a couple of months back of rewriting the front-end web layer of blogxter. It's now a bit long in the tooth (it was written before we had the scripting libraries). and was created more as a technology showcase than anything. We did stuff like having every PNG image be dynamically generated from SVG on demand. I want to introduce some more up-to-date design patterns and improve efficieny. One thing in particular is to take out the unnecessary and somewhat heavyweight use of XForms in the 'myblogs' tools - I'd like to support an additional BBCode editing mode and add dynamic AJAX previews. Anyone interested in helping?

Spurred into action following a recent meeting with the boys from Delta XML, we have created a live demo of an out-of-the-box NetKernel system.

The demo is a real live NetKernel instance and it allows you to explore the management tools, system status reports as well as the developer tools and docs. For security we've locked down critical parts of the system but most applications are available to try.

The technically interesting part is that the demo uses a functional filter-layer applied over the standard backend admin system - kind of like a see-through sneeze screen over a deli-counter. The implementation is a simple mapper pattern - all requests in the backend fulcrum are redirected to a mapper script. This performs a URI validation against a precompiled (transrepted) declarative access control list. If the requested URI is permitted then the request is transparently reissued, if not then a lock down message is displayed. Result is that the end-user doesn't know the filter is in place unless they stray from the permitted path.

Bytecode manipulating AOP eat your heart out ;-)

Jon has posted the podcast of an interview we recorded last week for his 'Interviews with Innovators' column at Infoworld.

He's done a great job of editing our 2hr conversation down to a tight 30mins. The discussion covered a lot of ground. Towards the end Jon revealed that he has a deep and intuitive grasp of URI-based software composition - hence this comment he makes in the discussion "Personally I'm fascinated by it [NetKernel]".

Gotta work on a few more demos to capture the essence of NetKernel. As I said to Randy recently, it feels like I'm Walter Raleigh just returned to London and enthusing about the wonders of the North American contitent and I just have a potato and some dried tobacco leaves to represent the new world!

Continuing this analogy, one statement I made that didn't make the edit: "NetKernel is a whole new world, we need a big sign hanging out saying 'Explorers Wanted'".

Brian Sletten contacted me yesterday to discuss some of the details of the performance analysis guide we provide for system admins.

In particular he asked about the relationship between throughput and response time in a loaded server...

Here's my reply to him:

"Throughput (requests per min, red) grows until CPU is saturated and then remains constant - the CPU is operating at max throughput. Once CPU is maxed out then response time must increase with load. In our case it increases predictably linearly with load. Essentially this shows NK is very close to an ideal linear system. A thread-per-request system would droop off as load increases since the CPU has to spend more and more time context switching threads and not doing real work."