UKOUG 2006: Day One
So I arrived in Birmingham with only hand luggage. There's further bad news: the exhibition hall was closed on Tuesday which means I won't be able to get through the week by wearing promotional T-Shirts. I'm going to have to go shopping. Besides, I don't expect any vendors will be giving away promotional underpants or socks. Maybe they are missing a trick here. I know people who would wear socks with a cartoon toad on them.
On the other hand at least I didn't have to check into my hotel immediately, so I caught almost the whole of Tom Kyte's keynote, What's Coming Next? This was a whiz through the likely contents of Oracle 11g (subject, of course, to the standard BETA! caveats). Tom devoted quite some time to the version control functionality, which he regards as the killer feature. It does sound quite neat, having an API layer of version controlled views, packages, etc so we can apply patches to our live applications without downtime. Certainly this might have saved me some late hours last month.
There are other features which catch the eye (such as native PL/SQL and Java compilation, virtual columns, flashback data archive) but I remain unconvinced that there is a compelling argument for upgrading. In John King's presentation on XML he did the traditional database version survey: some people on 8i, some more on 10 but roughly two-thirds on 9i. He said he came across the same proportions in the USA. Even Oracle only claim a 50:50 split between 9i and 10g. Oracle need to shift more of their customers onto newer versions for obvious business reasons. But as the major features in 10g (like the automated management stuff) still haven't achieved that it seems likely that Oracle will have to resort to coercion. At the very least they should seriously consider backporting some of the upgrade smoothing features to make it easier to move sites from 9i straight to 11g.
I chaired two sessions yesterday. This amounts to little more than reminding people to switch off their mobile phones and asking them to fill in the evaluation sheets. And giving the speaker time checks at the ten, five and one minute to go points.
Not that this was necessary for the first session. Chris Roderick from CERN spoke on Capturing, Storing and Using Time-Series Data for the World's Largest Scientific Instrument. Maintaining data from the Large Hadron Collider is an interesting business problem: sixty millions records per day (and likely to increase) that have to be searchable online, across a data set which will eventually hold twenty years' worth of data. The LHC Logging architecture is probably as close to a template solution as it is possible to have with such an esoteric scenario. They have an API which batches up the client records to allow bulk DML operations. The data is stored in range-partitioned index organized tables. The queries make extensive use of analytic functions. I would liked to hear more on the decision making process. How did they decide on index organized tables rather than regular ones? How did they model the scalability of the solution? But the UKOUG sessions are only forty-five minutes long and Chris did a very good job of explaining the project in the allotted time.
The next session I chaired went right to the wire. John King's gave us an overview of XML in Oracle. This was a (coherent) gallop through the entire feature set from simple basic SQL support through to WebDAV. It was like having the entire XML DB guide distilled and injected straight into your veins. John thinks that XML DB is primarily a marketing thing, gathering up a whole bunch of features under a name than is trademarkable. Oracle indulging in marketing? Surely not. One other he said that struck me was that XML in the database is used primarily for backup, in case people need to reproduce XML documents. I would like to think this is true but I suspect all sorts of nasty implementations are lurking out there.
In the film of The Right Stuff (and possibly in real life too), the first Mercury astronaut Alan Shepherd sits in the capsule on the launch pad and recites the following prayer: "Dear Lord, please don't let me f**k up". This is a great prayer for presenters too. I skipped lunch to run through my presentation one more time. Which was just as well because I discovered that I had left a whole bunch of DBMS_OUTPUT messages in one of my demo scripts.
From my side of the podium I think the actual presentation went well. The scripts ran, my mind didn't go blank and the questions at the end suggested that at least some people had been paying attention. There was one minor blip: when I submitted the paper it turn's out I had called it At Last! A Use For Objects whereas my slides used TYPE, because I was speaking about programming with SQL TYPEs. Nobody seemed bothered, but you can never tell: some OO people get terribly particular about terminology.
We're S-H-O-P-P-I-N-G, we're shopping
The bloggers' meeting
After the SIG Chairs' meeting we decamped to All Bar One for a joint SIG Chairs/Bloggers thrash. I discussed PowerPoint technique with Peter Scott. I'm moving towards a style that has fewer slides and fewer - or no - words on the slides I do have. I think Peter thought I was criticising him for having 55 slides in his session but I wasn't. Tim Hall is still jetlagged, poor guy. At least he's got his luggage back, whichI hope is a good omen for me. I completely failed to engage Babette Turner-Underwood in an interesting conversation about Data Pump. But she forgave me and we moved on to Funny Things Children Say. And Mike Durran asked me a very pertinent question: "So what do you blog about?" Mike, I'll get back to you when I find out.