I'll stay out of the discussion of front end and middle tier discussion other than to note that the technologies used for those architectures are in a state of flux. It's hard to tell what the "best" product is going to be in 5 years, and even the question of what will still be "popular" is hard to answer.
On the back end, though, the choices are pretty clear, and are likely to be stable for years. Here's my take:
- MySQL is actually a collection of products of various degrees of sophistication and cost. The product most people mean by "MySQL" is not a DBMS, but a file manager, and is not appropriate for a system as complex as MySpace. The Cadillac MySQL product, InnoDB, was robust enough to run Yahoo (while it was small), but was bought by Oracle Corp and is now proprietary.
- PostgreSQL is the frontrunner in Open Source DBMSes. It has the most DBMS features of any free ones, and is modeled on Oracle, which gives you an upgrade path and a huge labor pool to draw on.
- Some other free or Open Source DBMSes are interesting and may be attractive to niche markets. Firebird looks the most cool to me.
- Oracle is where you'll end up if you want a largish (500+ tables/ 100m+ rows) database and do a competitive analysis based on total cost of ownership. Most developers will argue about that, but trust me :)
When you're pricing technology, don't forget to factor in labor. That is why, for example, the expensive DBMSes can be cheaper than the free ones. If the only Firebird DBAs cost $200/hour and have to be flown in from Silicon Valley, you probably wanted something else.