As an SME consulting company, we often come up against large offshore development set-ups and the classic accountancy argument “We’ll use them as their day rates are a fraction of yours”.  There is an obvious problem with this – software development is complicated and expertise gained over years of coding, integrating and testing can lead to orders of magnitude of improvement in speed of development and subsequent quality.

We’ve recently come up against an excellent example of this.  While on client site, we were integrating to a credit card provider in Mule via web services.  Mule supports CXF, Axis 1 and Axis 2 as Java WS frameworks, and they all have positives and negatives so we advised the client to use the same framework as their offshore supplier had used in the back-office system to ensure support and maintenance was made easier.  We subsequently discovered that, as they had no experience of Web Service integration, the offshore supplier had used HTTPConnection and DOM – i.e. they were hard-coding each web service call.

That afternoon, we integrated all 5 webservices and used Mule’s definition XML to model the control process.  In 4 hours work we had acheived the equivalent of over 200 man days of offshore development.

Expertise was obviously the clear winner here!

One of the things that is a constant surprise is finding clients with slow Java systems that spend a tonne of cash on new hardware but don’t configure their systems to make full use of it.  A classic problem is leaving the application server settings as per first install, with very low JVM heap size and a poor garbage collector, another problem being running only one instance of a JVM, a situation where you may have a new server with 4GB of RAM and a quad-core CPU but running the software in only 512MB RAM with threading likely to be tied to a single core.

Sceneric consultants are experts at optimising our clients platforms to ensure the hardware is properly utilised.  Our consultants have deployed applications on clustered IBM Websphere, Weblogic, Oracle iAS and JBoss servers, and have created Coldfusion Enterprise clusters on JRun 4.0 leading to significant performance and stability gains.

The key thing to remember is that Java runs within a virtual machine which has limits ordinarily significantly smaller than the avialable hardware.   Without explicit tuning expertise applied, the applications running within the virutal machine will not be performing to their optimum potential.

Some things in life are just the “right” thing to do.  Take implementing a content managed website.  In the early days of content management systems (CMS), there was often no division of content from presentation.  This meant that changing branding & look and feel or re-using content across trading divisions was difficult and costly.  Also, the content was only available as HTML, there was no ability to make the content available through any other mechanism without considerable extra work.

However, even these systems could be implemented correctly, separating content from presentation and utilising content tagging to allow web experiences to be personalised.  Sceneric has built a number of award winning websites forcing this separation and providing the content as discrete, tagged items of information.  This very much the building blocks of the semantic web and can be demonstrated by Friends Provident’s news items being made available as an RSS feed with only 1 hours development, the first FTSE100 company to offer this service. Sceneric have published these guidelines as best practice which can be downloaded here.

This approach has been formalised by all modern CMS projects, in the JSR170 Java standard as used by Magnolia, LiveRay, Alfresco and Oracle CMS, and in the templated approach of Joomla, Drupal, CMS Made Simple in the PHP and Python arena.  In fact, Microsoft’s Sharepoint is also an excellent implementation of this idea.

The key point is that project implementation can be just as rapid and cost the same to produce a platform that seperates content and presentation as to rush something into production that will cost orders of magnitude more to fix in future.