The problem is primarily one of mindset. The presence of multiple applications at multiple versions and multiple environments on multiple machines can lead to a belief that it’s necessary to explicitly manage all of these axes as part of the deployment process.

I prefer a different metaphor: I always try to work as if the deployment artifact will be burned onto DVD and shipped all over the world. Treating software as a product like this makes it much easier to identify what is really part of the software and what must be configured as part of the installation process.

All of the environment/machine/etc specific information can be collapsed into configuration, eg. as a property file or command-line argument. Artificially separating them out leads to more complexity not less. It’s either part of the software (on the DVD) or it’s configuration. That’s it.

• Be written using node.js, Ømq, redis and riak.

• Be built with gradle
• Use git for SCM
• Check in deployable artifacts
• Be an executable jar
• Use JPA annotations for persistence

Would not…

• Use maven
• Use spring

After I stop cringing at how naive and earnest we all were I try and contrast my current opinions with those of my younger self and his young friends.

Some thoughts from tonight’s history lesson:

• Be aware of your biases, preconceptions and beliefs. Consider how they might colour your interpretation of events.
• Throwing away an acceptance test suite representing ten person-years of work because you want to use a different test tool is almost always a shockingly irresponsible thing to do.
• A major high street bank based in Poole could have had an award-winning business banking website if they hadn’t (ostensibly) cancelled the project because it was going to deliver 3 months late. That was 5 years ago. It’s still not finished.
• Technical excellence is meaningless without effective governance.
• High performing teams are messy, loud, exhausting, chaotic fun and will still be discussed fondly whenever 3 or more members get together, even 5 years later.

NOsql, Dynamic, Event driven, Distributed.

A tipping point has been reached. The JVM is past its peak. Dynamic languages with a low cost of change and greater expressive power are in the ascendance. Scaling through multithreading has been eclipsed by non-blocking single-threaded, multi-process architectures. Centralised databases are being out-evolved by distributed, eventually-consistent, semi-structured storage. Commodity hardware and designing failure-tolerance into the fabric of the system are winning out over specialised, high-cost solutions.

I’m seeing a lot of innovation in this area. It feels to me a lot like the ajax explosion. Before the term was coined, there were a lot of us ‘doing ajax’ without realising what it was we’d got our hands on, or just how amazing it could be. The XMLHttpRequest object was relegated to the back of the O’Reilly Javascript book and the most comprehensive documentation was a single page on Apple’s website. Once Jesse James Garett gave it a name, ajax was suddenly everywhere.

Thus: noded. A noded system is, generally: built on commodity hardware, horizontally scalable, fault-tolerant, (usually) in a dynamic language and operates at a scale far larger than any traditional solution.

So who has a noded architecture? Among others, Google (Bigtable, MapReduce etc…), Facebook (Cassandra), Twitter (FlockDb, Gizzard), Amazon (Dynamo). You may have heard of them.

A number of impressive open-source projects have been released by these organisations and others, and more have been inspired by them: Hadoop, Riak, MongoDB, CouchDB to name a few.

The platform that has me most exited at the moment is node. Javascript is a hugely expressive language (even more so with tools such as CoffeeScript) and node is rapidly accumulating an impressive collection of libraries that make building highly efficient, ultra-scalable applications incredibly easy. It’s not a coincidence that I like the term noded. I’d also settle for ‘nodular’ but Charles claimed that one in 2003.

The slideshow

Finding myself once again spending the christmas-new year limbo period mucking about with a year’s worth of backlogged blog administration and semi-traditional experimentation with alternative hosting. I wouldn’t mind except it’s almost impossible to avoid reading stuff I wrote 8 years ago and cringeing. I think I knew more then than I did now. Or perhaps was just more enthusastic about broadcasting it.

Although I love your products, as long as you continue to pursue your lawsuit against HTC, I shall no longer be buying any of them for the forseeable future.

Best regards,

Darren