I would define the project first, then consider my alternatives for building it. Perhaps I would not have to “build” it at all, as in “from scratch,” but would find an already-existing framework that I could just drop into place.

Let's face it:   by now, most web-sites are very routine things, and high-level construction tools are beginning to emerge as “truly battle-tested.” That's what you want... something that somebody else has already invested a lot of time and hair-follicles into, so that you don't.

When it came down to the actual implementation, it really would not matter much. I've done 'em in Perl. I've done 'em in PHP. I've done 'em in Python. Usually there was something in the way of already-installed software or already-in-house expertise, and I simply adapted the implementation-strategy to that consideration.

For any web-project, you're really only interested in the results obtained, and the fact that you're done for the day, and in the pub sipping from a tall foamy mug, at a reasonable hour and without regrets. The last thing you want to find yourself doing is “writing yet-another program.”