Selecting a web framework can be a somewhat daunting task. Its not too bad if what you're building is a one-off, standalone application that isn't too big. Unless you need to do something very left-field, every half decent framework will do at least "fine" or better. There will be forms and databases and validation and html views. You know the drill.

But what about a long term project that you and your colleages need to continually develop for until the end of time (or rather, when the next rebuild happens or when you're all out of a job)? Clearly it is foolhardy to jump on the bandwagon of the latest cool-kid in framework-town. Longevity now counts for a lot. You need not only to do the standard stuff and do it well, but you need to get a feel for how you will deal with problems that you haven't been presented with -- though plenty of motivating examples are very useful -- and someone will always ask for something weird.

I think this raises some interesting points about making decisions more generally, specifically about technical requirements. It would be awesome if we could make this choice without having to fall back on instinct. I'm not saying that instinct is all you have -- you can quickly start to rule out non-contenders that are clearly well off the pace -- but unless you are (by some measure) very fortunate, several will look like a good bet, and the things that will differentiate them are only a long way down the line, and only the very lucky can give the options a full road test.

And this is the position we are currently in: Symfony vs Zend vs Phalcon. Symfony wins at modularity and features, Zend wins at business support and Phalcon wins at performance. Which one will it turn out that we really do need most? Its not that we require at least "seven" performance and Symfony only has "five", how much is enough? Nor even which features are available -- we can always build a feature if it doesn't exist, but will the time lost be worth it versus the gains we made by eschewing the possibility of having it already

In a domain where we feel so comfortable knowing that our decisions are often so strongly evidence-based, when making a decision that will impact on the next how-many-years of development, it feels so unnerving to let gut-feeling have any sway. Though there's comfort in the fact that, for our current quandary at least, every option would be a massive improvement.

Except Zend.