Whoever is in charge of software architecture, be they senior developers, whole teams like in agile practice or architects-per-se, it is a deep trend to apply what I like to call Ego Driven Architecture (or EDA for short, not to be mistaken with Event Driven Architecture).
When one has to choose an architecture, one should design it from a number of objective criteria, including:
- business requirements,
- technical constraints,
- ease of use,
- maintenance costs,
- etc.
One could even argue you should take care of subjective yet real constraints:
- cross-service warfare (between the development team and the production team),
- interpersonal problems (between a project manager and the lead developer),
- historical bias (not using EJB3 because EJB2 where too complex to develop),
- etc.
Normally, each criterion is then assigned a priority, and the designed architecture’s objective is to answer the most criteria, based on their priority. Alas, when using EDA, the top priority criterion is completely different: whatever the constraints and the requirements, the technologies used on the project should make the CV of the architect(s) shine brighter.
If you already don’t know what I speak of, imagine an architect using EJB3 on a project though the production application server is not ready to run them yet. Or an architect pressing the production team to upgrade to the latest version of the application server, although there’s no real need to. I know you’ve already seen such behaviours.
There’re a number of factors that contribute to the thriving of EDA:
- There’s no validation of projects architecture by a dedicated technical team. Either there’s no such cross-project team or it has no influence over architectural choices, due to the business funding (read power). Remember that since the money comes from the business, it is always very hard to tell them 'no'.
- Architects in charge of the design are junior. They definitely want to put to good use what they read in articles posted on the Web. Not to offense anyone, but before using Scala and such, wouldn’t it be better if someone already had enough feedbacks?
- Architects are only hired for the time of the project. If your contribution stops when the project is shipped, it is a very big incentive to try unconventional technologies. Try talking with internal IS teams and you’ll find the 'veni, vedi, vici' syndrom a top resentment cause for outsourcing since they usually end up cleaning the crap.
Still, the root cause of EDA is a miscalculation on the architect’s part. The IS world, even globalized, is very very finite. You always work with the same persons. If you practice EDA, sooner or later, you’ll find someone who will remember you for being the one who proposed the architecture who killed the project. Our work is complex enough taking every business requirement, every technical constraint, every service warfare into account without bringing the architect’s ego into the fray. So please, refrain from using EDA
This doesn’t mean you shouldn’t strive to use pertinent technologies in your architecture but it should be based on objective reasons, not you carreer.