After using mongodb for quite some time and building several systems with it, I am starting to drift back into preferring structured data in a mysql db. 

• It is reliably fast
• Building the structure of the mysql seems to help in the designing of a system
• It seems more stable

I think the second one is the biggest reason. It is a rather fuzzy thing, but I like to look at the db and see what fields are there. 

My convictions are weak here, and I will probably drift back at some later date, but I just feel that the relational db model matches well with how a web based system is designed.

Along with that, I am drifting back to PHP for web systems rather than node.js. There just is less layers of things that you need to cobble together to put up a php web site, and less points of failure. 

Perhaps I am losing my hipness. I just like my web systems to be stable, fast, and easy to develop. Counting up the number of hours of work setting up and keeping stable mongodb/node.js sites, and excluding time spent just making things that actually spit out a novel piece of html for a browser, makes me reconsider these new shiny technologies. 

Now, with that all said, there are some cases where they really shine. Node really lets you get nitty-gritty with the type of server stuff you can do (such as real-time comet connections, or just plain old tcp sockets). Mongo allows you to really dump models right into your db - goodby ORMs (and good riddance)!

But, if you want to make a website where users make requests to the server for html, and you need to store some data about the users and the results of their actions, well PHP/MySQL is still the shortest path between A & B.