16 August 2013

This Is Madness

Two months ago my time with Big Blue ended. While preparing for my Journeyman tour next month, I sorted out my drafts of blog posts and rediscovered some notes concerning my last assignment in the company. It was a particularly crazy time and I only wrote about it to clear my mind. Nevertheless, such gems need to be shared. As usual all my writing is fictional and does not resemble any real people or companies.

For my latest assignment I joined another team. I was happy at first because I worked with a new prototype, which was a nice change from my previous waterfall project. The team was led by a Distinguished Engineer, so my expectations were high. But soon the forces of madness showed.

One Does Not Simply Write His Own DatabaseRoll Your Own
First I noticed that one developer had created his own in-memory document store to support OLAP style operations, mainly pre-aggregated sums. I was not sure why he had done that, he had not checked existing third-party products before, but I guess creating your own database is a fun project to play with. Unfortunately many concepts where missing and the provided generality was not needed. (Note: One Does Not Simply create his or her own database system during a development project.) As a new member I did not want to question him during my first weeks and just accepted it. Later when he wanted to add some features, he rewrote the whole thing from scratch during a weekend, dropping at least one month worth of my additions. I was not amused.

Web or Desktop Application?
The project was under active development for almost a year and the business stake holders had still not decided if they wanted a web or a standalone application. The core developers were used to Eclipse/RCP, so they had created a server-client application prototype. When the business was finally ready to decide, the whole team spent two days creating estimates for both variants. "Surprisingly" porting the already existing desktop client to the web was rejected as being more expensive. I guess the business did not really care. (Note: Architecture decisions need to be made early in the project.)

Upload a File
Users needed to upload a large file to the server which would take several minutes. It needed to be extra reliable (but the exact term of "extra" was never defined). So the whole team met online to discuss the issue. One developer had the idea to use messaging for the upload because "messaging is usually reliable" and the company's flagship JMS product was extra reliable. I could not believe it - WTF? The client would still have to upload the file to the server to put it into the message queue. Further messaging is not designed to work with huge files. I begged the lead developer not to introduce a new component only to transfer a several MB file from the servlet layer to the service layer. At this time I started sending emails with "Madness? No this is reliaaaaaaable".

Madness? No This Is ReliableFrom Prototype to Production Over Night
The project had started as research prototype with the goal to experiment and compare different algorithms. Some of the numbers did not make sense and were being discussed and changed by the business people every day. Suddenly things changed, the whole project became production software and a deadline came up. The deadline was - of course - completely unrealistic but nobody told the business. Further more, there was a strict protocol for production applications in the company which needed many documents and formal reviews. Creating these artefacts delayed the deadline, too.

The Grand Finale
For all this bureaucracy, a young developer was brought in as architect. He did most of the necessary paper work which really helped but then things got nasty. He introduced code reviews - which is a good idea in general - but his findings were useless. First he forced us to use a large JavaDoc template which introduced much noise into our classes and then he demanded that every method would log its entry and its exit. My heart sank, but fortunately my twitter followers gave me permission to ignore his rules ;-)

Again this is not much more than a rant, but it is also a true story. Some of my friends enjoy reading rants, maybe my stories reassure them that one should not work for large corporations.

No comments: