My favorite encounter (on behalf of the M.I.L. whose phone they called) was about 45 minutes. The pattern from previous calls is they want you to type "(win)+R" to then launch event viewer, because OMG all the errors they'll show you.

So I stymied them for over five minutes when I explained my keyboard didn't have that little windows key on it.
Oh, yes, I know what it looks like, my neighbor's PC has one, but I don't.

Then after some mention of "this thing is awfully slow, I'll have to restart it" (setting the phone down) was more minutes of his life gone. Then accidentally hung up half way through a sentence. They actually called back. Then allegedly the kid was having a meltdown, very sorry sir I have to check, but I'll be right back, and I do want these problems taken care of. Et cetera.

The jig was up after I "tried" to download their remote control software, but my (IE, allegedly) said site could not be found. (I was instead googling the error message to simulate a 'dialup modem not connected' scenario that they'd have to walk me through. You know, stay a step ahead.)

Finally Dave's "manager" cursed at me and said that I had ample time to get this to work and that I was wasting their time. A realization that was honestly 30 minutes too late. (Oh, no calls since.)

Was cold-called by someone from Google, so I entertained the invite to interview, partially on a whim. I agreed I would not discuss specifics of interview process or questions, which I'll abide by, by suffice to the theoretical "design a system that..." questions and "how would you..." questions I did fairly well on and felt rather confident about. The detail coding component ended up drawing a question I wasn't all that prepped for (not being something that for me has ever come up in long years of working). So feedback was that due to the coding question performance, it was a no-offer. Which was fine -- after finally seeing SV area in comparison to where I'm happy with in fly-over territory (partly the area, especially cost-of-housing), I probably wouldn't have accepted anyway. No harm, no foul; just not my thing. Still, it wasn't a terribly great assessment of what I know and bring to the table. Local market does the same thing. /shrug

My own team here, we tend to assign a homework question with semi-vague requirements, and then grade on what their experience taught them to also include, in addition to implementation specifics, with a follow-up Q&A ... this seems to work.

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 ... but maybe it speaks to a desire to push all our installs to their flagship DB.

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."

"Exactly."

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 ... they chose neither. With the writing on the wall, the guild fell apart, cannibalized by Rift and LOTRO. And the same story happened with some of the remaining 19/25ths of the progression raid crew, leaving stragglers like me with no team to run 10's or 25's with in Cata. I cancelled my account as much by attrition as disinterest in 30+ minute queues to re-run the same couple dungeons each day as a filler. (But we expect we'll regroup as soon as SW:TOR comes out, if it's even half decent.)

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.