One rather minor thing Oracle tells us they'll add to a future update of MySQL is to capture an important piece of data missing in the binary log files. BInlogs currently don't track which db user is issuing DDL or DML statements, so filtering out changes against this is not currently possible. Admittedly not a large use case scenario, but for us it is an "almost-blocker" for some replication work. (Our source systems need to occasionally purge material, and we don't want the purges replicated to the destination. Oracle sales sold us on doing filtering by the db user before we found out the binlogs don't even have that info.) Hopefully any open source fork (MariaDB or otherwise) picks this up or stays compatible.
I mentioned elsewhere (AC, alas) that our company, when it came time to re-license, found it cheaper to [decide to] move a couple/few hundred commercial MySQL installations to Oracle SE instead. This may be a "win" for us in the long term since the central data warehouse is itself Oracle EE. Maybe it speaks to a cash grab for the existing MySQL installs
And we're generally aware of MariaDB, but it didn't get much consideration.
"Look at the irony settings in our cover song... they all go to eleven."
"Oh I see. And most cover songs only go up to ten."
Something that contributed heavily to my PVE guild (which did casual 10-man raids for lulz) and my standing raid crew (25-man, progression-minded) was the raid lockout system change they made going just prior to launching Cata. Those two groups of raiders overlapped by about 6 people -- people who enjoyed playing the same toon in both settings. The "flexible" lockout system killed this arrangement: any toon could only kill one boss -- be it in 10-man or 25-man groups -- one time per week. (Fail.) So our powerhouse players could not also help out in the casual guild runs. That the progression raids sometimes opted to extend lock outs into the next week to keep working a particular boss also contributed. So with players feeling like they were being forced to choose which groups of friends they wanted to raid with in any given week
One of my accounts also got a message from Brookstone a day or two ago.
One of the big reasons is that HTML lacks semantic meaning beyond simple paragraph constructs. Documentation-oriented markup languages (of which I'm more familiar with DITA) and schemas can seem arbitrarily complicated to a casual observer, granted; but having an identifier that clarifies "this" paragraph being an instruction that should be executed by the user, and "that" paragraph being merely an example can allow for some rules-based (automated) processing to exist between authorship and production that wouldn't be possible lacking some notion of the semantic purposes of a random collection of raw paragraphs.