mult.ifario.us

I spent some time cleaning up in our yard on a recent weekend, and in the process, I ran across a good-sized spider that I didn't recognize. It had a leg span of around an inch and a half — good-sized by Pacific Northwest standards — and my ill-informed initial guess was that it was probably a harmless and beneficial wolf spider. I've been trying to get the kid interested and informed about spiders, snails, bees, yellow jackets, ants, beetles, turtles, frogs, and other animals we come across, both in terms of identification and in terms of information (don't touch this one, touch that one if you want, tell Dada if you see one of these, etc.). I showed the kid the spider, snapped a couple of pictures, explained that it was a look not touch type thing, and it eventually crept off under the fence. So far, so good, too — ever since she got over her irrational fear of bumblebees, she's been both respectful and interested in whatever creepy-crawlies we come across.

Based on my knowledge of the spiders around my home town of Portland, OR, around 200 miles south of here, I made the assumption that the most dangerous insect or arachnid in the area was the mean and aggressive yellow jacket. However, after I did some Googling and reading to figure out what kind of spider we'd found, I may need to be more careful. The hobo spider (Tegenaria agrestis) is becoming common in the Northwest and has a very nasty bite, at least anecdotally. At least one source claims that there is no strong medial evidence to support the assertions that the spider has necrotic venom, but I won't be conducting any experiments.

The unfortunate thing is that the hobo spider is indistinguishable from its otherwise harmless cousins the Common House Spider (Tegenaria domestica) and the aptly named Giant House Spider (Tegenaria duellica) other than by size or by examination under a microscope. I exchanged a couple of emails with a friendly arachnologist about the spiders, and he claimed that I'd never heard of or come across these spiders in my youth in Oregon because they were introduced in British Columbia some time ago and have only recently extended their range into Oregon.

It turns out that I've got a few of the duellica living in the yard, and they are big. They're not tarantula big, but they're at least double the size of any other species of spider that I've seen in the region. Out of curiosity, I baited some of the funnel webs around the yard with dead yellow jackets and then wiggled the web with a twig to see who came out to claim the free meal. In one case, a spider with a legspan the size of my palm and a body as thick as my pinky lumbered out, grabbed the snack, and carted it back under the window sill where his web was. I'm on the fence between making sure that the kid knows not to stick her finger in the web and relocating the spider, but I'm probably going to take the relocation option. (I'd prefer not to kill the spider; the duellica is good to have around, as it displaces and/or kills the agrestis.)

The 2007 Fremont Solstice Parade happened yesterday, and while it wasn't as creative as last year's (photo set), it was still interesting to see what the community came up with. The combination puppet and costume in the middle below was my favorite, but the photo doesn't really do justice to how natural the motions of the puppets were. The real disappointment this year was with the hooting and hollering from the crowd, where that kind of meathead behavior hadn't been as apparent the previous year in spite of an equal number of naked or effectively naked people parading around.

Here's a link to my 2007 set. The shots are cross-posted in three photo pools, but don't bother if you've got issues with nudity or liberal politics. (Here's a set with some good shots of the cyclists getting painted before their ride.)

As taking pictures goes, I took my D200 instead of the Leica Digilux 1 that I was using last year, and I continue to be infinitely happier with the DSLR than the point-and-shoot. The D200 has been particularly nice for the minimal power-up time and the ergonomics of the controls; I can dial-in a different ISO setting, bracket five shots across +/-2 stops, and squeeze off a pile of frames in a second, all without taking one hand off the kid sitting on my shoulders. (The "prosumer" features double as dad-friendly features...) While the Leica took good pictures and was my first digital camera, my previous camera was venerable a (non-digital) Minolta XD-11 that I'd inherited from a relative back in the early 1990s, and no amount of digital convenience could make me forget the level of control provided by the the SLR.

I recently came across a WSDL import statement of the following form:

<wsdl:import namespace="some://uri"
             location="../a/relative/path.wsdl" />

My first thought was that this is wrong, but I found out that it is actually standard behavior according to RFC 2396:

Within a relative-path reference, the complete path segments "." and ".." have special meanings: "the current hierarchy level" and "the level above this hierarchy level", respectively. Although this is very similar to their use within Unix-based filesystems to indicate directory levels, these path components are only considered special when resolving a relative-path reference to its absolute form (Section 5.2).

The RFC also specifies a sensible set of fallbacks for determining the base URI, rendered in ASCII art:

.----------------------------------------------------------.
|  .----------------------------------------------------.  |
|  |  .----------------------------------------------.  |  |
|  |  |  .----------------------------------------.  |  |  |
|  |  |  |  .----------------------------------.  |  |  |  |
|  |  |  |  |       <relative_reference>       |  |  |  |  |
|  |  |  |  `----------------------------------'  |  |  |  |
|  |  |  | (5.1.1) Base URI embedded in the       |  |  |  |
|  |  |  |         document's content             |  |  |  |
|  |  |  `----------------------------------------'  |  |  |
|  |  | (5.1.2) Base URI of the encapsulating entity |  |  |
|  |  |         (message, document, or none).        |  |  |
|  |  `----------------------------------------------'  |  |
|  | (5.1.3) URI used to retrieve the entity            |  |
|  `----------------------------------------------------'  |
| (5.1.4) Default Base URI is application-dependent        |
`----------------------------------------------------------'

Nonetheless, this can still be a bit confusing. In this case, the WSDL that I was looking at was in a source code download, and the referenced resources were not. The best practice should be to specify an xml:base in every WSDL document that uses a relative wsdl:import location. (The WS-I Basic Profile doesn't say anything about relative wsdl:import locations.)

Just saw this over on Thingamababy:

RC2 Corp. Recalls Various Thomas & Friends™ Wooden Railway Toys Due to Lead Poisoning Hazard [...] Manufactured in: China [...]

While this is not necessarily an example of the same kind of wanton (wonton?) disregard for health and well-being as melamine in pet food and the ensuing food safety debacle, there won't be any more Chinese-made toys (or food or anything else) in our house until we've got a good reason to believe that their safety standards are significantly higher. As I understand it, lead paint dust is the major hazard as opposed to skin contact with intact lead paint, so the kid probably consumed very little if any lead playing with the toy trains. But that's not the point.

This is a failure on the part of the distributor to properly monitor and police their suppliers — someone at some level at some point knew better. Class action anyone?

My cellular service contract was coming up for renewal this month, and I'd been thinking about buying an iPhone as an experiment. After thinking it over and without waiting for WWDC to happen, I bought a BlackBerry 8800 instead, and this fragment from the WWDC "Jobsnote" transcript (gone missing, so no link) on MacRumors explains my reasoning perfectly:

11:22 am, Head of Apple iPhone software had some trouble typing with the iPhone on-screen keyboard.

With all of the various mobile devices and phones that I've used (since the Newton was unable to decipher my handwriting), I've never thought that the UI wasn't pretty enough or "If only I could watch a tiny movie right now..." I've always hated the user interface and data synchronization experience.

So far, I'm pleased as punch with the Blackberry. The QWERTY-esqe (QWERTY for letters, modifier keys for numbers and symbols) keyboard and scroll nub is the only usable user interface that I've ever used, and that includes touch screen keyboards accessed with a stylus on PDA devices. The application stack is also a win for me, since there's direct support for GMail, Google Talk, Google Maps (the 8800 has GPS functionality), and the combination of Missing Sync and Spanning Sync synchronizes the Blackberry's calendar with my Google Calendar with my iCal on my various Macs. (I opted for Missing Sync over PocketMac after talking with some other Blackberry/Mac users about their synchronization experiences.)

The one open item is whether and how to push my open items from Remember The Milk, since the mobile interface isn't that usable (IMHO), but I'll probably use the offline support via Google Gears to just leave that laptop-only functionality.

All of the various user interfaces (web, mobile, console) for the different components and the multi-hop synchronization will make an interesting case study in whether this works well from a user perspective.

A consulting customer of mine who deploys Java apps on Windows had an issue with an otherwise well-behaved system that was suddenly thrashing. A quick glance at the virtual sizes of various processes showed an instance of winlogon that had sucked up over 2GB of memory was the culprit, and the underlying reason was amusing:

[...] If you do not log on the computer, a system memory leak may occur. Over time, the operating system may leak memory in the Winlogon.exe process at a rate of approximately 1 to 2 megabytes (MB) an hour. [...]

At that rate, ~2GB of leakage meant that the system had been up for between one and three months, which is not so good.

But then maybe it's not really so bad, either. My choice of a deployment platform would be different (Solaris or Linux), but that's more out of familiarity and personal experience than any hard data. Realistically, rebooting once a month shouldn't be a big deal. A reboot is a dry run for a recovery scenario, so bringing a machine down and back up should be routine. In an N+1 configuration, a rolling reboot shouldn't impact the customer, either.

In any case, I don't think that Windows Server 2003 was designed with recovery-oriented computing as one of the guiding principles.

I sometimes get to thinking that from the Java side of a layer like Hibernate or OpenJPA a database is just a database, but that's never quite true. One example is date/time granularity. The SQL standard specifies that TIME or TIMESTAMP datatypes can support fractional seconds but doesn't require it, and different databases do different things. For example, MySQL drops the fractional component of seconds altogether, while SQLServer rounds to the nearest 333rd of a second ("increments of .000, .003, or .007 seconds").

The SQLServer behavior manifested itself as a Heisenbug in a lightweight database-backed caching algorithm. An update timestamp was maintained by the application in the form of a Java java.util.Date, which has millisecond precision. The update timestamp was passed to newly created or updated objects and then round-tripped to a SQLServer instance in the course of an update pass, and the cache was throwing out all entries about one-third of the time because rounding of the round-tripped value made the entries look old to a numeric comparison of the long within the java.util.Date. Adding a second to the object timestamp before comparing solves the issue.