O2Active
23.3.2007
Background: When we were approached by the company to redesign the portal, we requested specification of what they were looking for. Apart from a series of web templates, they had no specification save that it would be expected that site traffic would be at 650,000 hits per month, need to be modified by internal and external employees, refresh its content on certain intervals, be updatable by partners whose content was on the site, it should work, and not fail. In response to this, I proposed to the company a design based on some of my academic ideas using C#.NET (VS2005), SQL2005, XML, XSL, XPath, and XSD (Altova). The idea was to become a unique, secure, and well-adapted content management system (CMS) where employees could log in and make changes to content without being exposed to the coding environment. At the software’s core, I used a design called an Extended Kohonen Map which is a form of an artificial neural network that could self-realize the loads placed on the system at any given time and self-construct the dynamic pages. The algorithm is quite robust when using associative memory in applications and services and I was very confident the notion could be put into practice on such a large scale.
Implementation: the course of the next six weeks, I set out with a team of seven employees to build the portal. As architect and lead developer, I managed the project by dividing into small teams: 1) the site manager protocol (REDS), 2) Core.dll, 3) xml assimilator, and 4) response-write. Coordination was simple as we all worked in the same office but I found some pitfalls in having to educate the crew about my passionate vision of what the idea was, how the nodes should be balanced to insure integrity, and general difficulties in getting people on track about how the XSLs should function in tandem with the incoming XML streams from the content providers. There was also the problem of a language barrier as I did not speak Czech well enough. We had additional difficulty in getting the portal’s partners to accept our new formatted XML/XSD system embedded in the required SQL tables. After a few meetings and using my charm, I was able to win them over. However, I learned that I needed to bring my language down from a high academic level to a more everyday tongue of the developers and to non-technical people.
Result: Happily, I can say we launched on time, on budget, and to date since 1 September 2006 the site has not experienced one crash. This is due to the robustness of BPS or Boagaphish Protocol System. As a reward, Ahoi was treated to much fanfare and publicity and became a Microsoft Certified Partner for our implementation. The full site can be viewed here.
CreditInfo
23.2.2007
Background: Credit Info was an Icelandic financial house that managed credit records of businesses across Europe. They assembled credit reports for companies throughout the world and, for a service fee, made them available to interested parties. They had been quite successful business-wise so were interested in attracting clients from the former Soviet republics. They had a system to manage the records coming in from various banks across the world coded in XML but had been experiencing failure in their system importing documents into their server which was frustrating many of their customers. They asked for a redesign motif and I presented one called datapump.
Implementation: Datapump was based on a design I prototyped called The Assimilator. Inspired by the Star Trek concept of the Borg—a habit I have using references from sci-fi when naming applications—I constructed the prototype and presented it to my superiors. The reaction was positive and they asked if I could put in into service. I managed a small team of two and reported to a senior developer about our progress. In the span of two months, we had the Assimilator (VS2003) running at a fast enough speed where it could put the incoming XML streams, validated by a XSD schema, into SQL2000 tables. It was able to digest 20TB of data in a little under 17 minutes without crashing and able to report its result as a log file sent to managers or employees responsible for deploying the data on the webservice. Some pitfalls were coordination with employees regarding the methodology in place, however, I created an intranet Wiki where I reported our progress and presented tutorials of the philosophy of the design and how it should be implemented. The “Wiki Solution” was a good solution to the problem of coordinating tasks of an unknown nature across a webservice and served as the model to train future employees about the software, how to use it, how it was constructed, the vision behind its design, and from what I understand, is still being added to.
Result: The project launched and was well received by upper management. The owner himself came in from Reykjavik to monitor the final stages of our implementation. It was extremely gratifying to have him shake our hands and say “good job” upon completion.
Monster
23.12.2006
Background: Monster is an online resume bulletin-board where users can post their CVs for free and employers can create paid accounts to review them. After working as maintenance of the core in VB6 (was one of three employees charged with the task), the company was looking to upgrade it to a newer .NET2.0 technology. The Monster site was quite large and difficult to ascertain as the code groups were not properly commented. It was a nightmare navigating through it trying to find particular failures as they arose.
Implementation: One of the employees in my group proposed a redesign idea building parts of the .NET framework and deploying class libraries in sets at a time. Following his instruction, I aided him in fulfilling his vision of how the site should be organized. We had several team meetings, coordinated with team members outside the office and across the United States and Europe to get everybody set on how the new site should look, behave, and how code was written. I particularly set up the procedures for commenting the code and wrote documentation regarding the philosophy and how the procedure should be employed. Pitfalls included miscommunications with others, not understanding why we were doing what we were, wondering why we were doing it at all, and how the final product should look.
Result: About 60% of the site was recoded before work terminated. Upper management revealed that the task was too great and we were instructed to stop our efforts and return the site to its state before we started the project. The reasons were unknown to me but it was rumored to be more office political than practical. The employee who authored the idea left shortly after and I resumed the maintenance of the old VB6 code.
Personal Application Projects--The RSS Reader Beta
21.2.2007
I have been playing around with what can be done with C#. Here is a stand-alone application. I will make the code available for download here shortly.
