Silicon Valley Redneck

Need to start wearing a neck gaiter

No, nor for politics nor Pestilence.  And certainly not in public.

I need something to keep grass clippings from getting down my shirt.  And maybe I should cover my schnoz with it, too; might cut down on the sneezing.

I've mentioned this in comments elsewhere

As we see more specifics about the members of the mob that's busily burning cities in the name of Social Justice, something kinda springs out at me.

Seems like "Anti"Fa is recruiting from pretty much the same pool as the Sturmabteilung - the actual, literal Brownshirts.  Well, no surprise there.  Organize a sociopathic, violent movement, attract violent sociopaths.

But then... we also hear high-visibility Democrats egging the Mob on, encouraging attacks on Republican public figures and random members of the public suspected of Republican sympathies.  And what we don't hear is the Democrat leadership denouncing this.  So... functionally Sturmabteilung?  Like, quasi-official paramilitary wing of the Party?  (Note that, at least in Blue-bubble zones, Democrats have privately been speaking of Republicans as subhuman and unworthy of life at least since the late 1970s, and this sort of speech became perfectly acceptable in public after the 2000 election.  It's not about Trump; the "basket of deplorables" thought, if not the wording, goes way back.)

I'm developing a nasty suspicion that, assuming there are actual polling places come November (so that people can cast secret ballots in the privacy of voting booths), there will be groups of masked, black-clad, menacing-looking individuals "helping to provide security" at polling places in many areas.  Possibly pointing their cellphone cameras at approaching would-be voters.  And maybe those phones will have a facial-recognition app, which maybe will communicate with a database of known unreliables and suspected Republicans.

For those marking their ballots at home, well: social pressure.  The family that votes together votes the way that won't get them yelled at and banished, right?  And expect a lot of pressure for people to post photos of their marked ballots on "social" media (there's a reason for the laws against this).

Ugly times indeed - and, in retrospect, the ugliness was already building long ago, back when I was living inside the Blue bubble.

Update:  Here's an instance of a Democrat (at the State level) standing up and arguing powerfully against the Mob.  But... seems that even before this speech, the Democrats had kicked him out of the Party for wandering off the reservation.

Ah, computers....

So now that I have that HDMI KVM switch under my desk, connecting the monitor to any of three computers plus a spare port, and it has an audio-out jack, maybe it's time to switch from built-in analog audio to HDMI audio so that'll switch too?

Start playing around with the mixer settings / audio configuration on the old workstation (since that's still the primary computer at this point).  Set everything for HDMI being the preferred destination.  Use the test buttons in the KDE mixer application.  Seems good.

Fire up Chrome; try watching a YouTube vid.  No audio.

Try playing an audio stream on VLC.  Audio.

Futz around some more.  Reboot.  Now various dedicated media players (not just VLC) work.  Neither Chrome nor Firefox has audio.

Looks like something in the venerable software configuration (and accumulated configuration cruft) can't make up its mind about which HDMI audio output device is real at any given moment, and trying to play audio  from a browser causes it to change its mind.

Oh, well.  I guess that's a further incentive to migrate to the new workstation, isn't it?  Meanwhile, I can swap the speaker cable between the old workstation and the KVM switch as needed.  Also, I need to reconfigure the audio setup on at least one other household computer to use HDMI instead of built-in analog, but that's with a recent install of a current distro.

Update:  Ordered a rudimentary stereo mixer from Amazon to put under my desk so's I won't need to move the plug around.  (I think I already have one of those around here... somewhere... in some uninformatively-marked moving box out in the barn.)  The machine in the TV room didn't think HDMI audio was a thing until I rebooted it, after which it worked fine.  (It had been needing a reboot anyway, having developed a case of absent-mindedness about how WiFi works.)

Update 2:  It seems that the TV-room machine forgets that HDMI audio exists when it goes to sleep and wakes up again, or maybe when the projector is turned off, and whichever it is, the presence of a powered-up projector on the HDMI port doesn't cause the HDMI option to reappear in the audio setup, never mind remain the output-in-use.  So, we may need an analog mixer there, too, to combine the computer's analog output with the projector's output (which will be active when using the Blurray player, coming soon - no, I'm not putting a priority on configuring things to be able to play and/or rip Burray disks under Linux).

Arguing with the tools again

So I'm back to playing with the TI development tools for the CC2640R2, or at least trying to.

Ended up deinstalling both versions of Code Composer Studio, and reinstalling just version 9 (the older one), as root this time, so it'd set up the udev rules for the debug pod.

Then, I tried running one of the example projects.  This resulted in an error, but at least a different one this time: the debug pod's firmware needed updating.  So I told it to go ahead.

And, error.  The error message suggested resetting the device and trying again, so I did.  Rinse, repeat.  After a few iterations, I decided (against all instructions) to try resetting the bugger by pulling the cable and replugging.  Oops!  Now I get a different error, suggesting that the firmware is now inoperative in some maybe-permanent way.

Grrrrr... maybe scratch one LunchPod, er, LaunchPad.  Well, I got three of them, and if I can get one of them working, I can use that one's debug-pod section to program the target devices on the other two.  But before I try anything, I'd kinda like to know what went wrong this time, so I don't end up scrambling all three of them.

Maybe... maybe... it's somehow related to my having had /dev/ttyACM0 open in gtkterm when first attempting the firmware update.  Maybe?

(Gee.  I could buy a standalone XDS110 pod for about $100.  And then probably scramble its firmware exactly the same way.)

Update:  For whatever reason, the pod just wasn't wanting to work until I rebooted the workstation.  Well, there were some out-of-sync kernel modules from the last update, so it was past time for a reboot.  Anyway, I successfully loaded and ran Project Zero... but then the target stopped responding to attempts to reprogram it.  Argh.  I seem to recall getting into this situation with some other part, maybe a Kinetis, with an app that disabled the debug interface in some way.  More investigation is called for, but at least the debug-pod side of the board is still in working order.

Update 2:  I'm getting that distinct impression that the combination of CC2640R2 LunchPod, Code Composter v9, the CC2640R2 SDK's demo projects, and the XDS110 firmware adds up to Not Ready For Prime Time.  The device having once been programmed via the debugger, it can't be reprogrammed until it's been erased, for which purpose TI's own Uniflash utility works just fine using the same embedded XDS110.  And, launching a demo app via the debugger doesn't appear to work properly, or at least the board doesn't show up on my phone's Bluetooth device list until I push the RESET button on the board, which terminally confuses the debugger.  Maybe it'd work better with CCS v10, except that v10 didn't (when I tried it) play nice with the CC2640R2 SDK.

Update 3:  OK, so I tried CCS v10 again... and it turns out the problem isn't an incompatibility in a tools bundle, as I'd thought, but rather a change in a library function bundled with TI's C compiler.  Applying the suggested fix to osal.c in one of the demo projects appears to solve the problem.  And!  It looks like the debugger in v10 is behaving itself.  So, time to play with some of the demo firmware, get a feeling for how it works and what the hardware is capable of, and then see about hacking up some application firmware for my own purposes.

Oh, and looking at the CC13xx/CC26xx manual?  Still with the brain-dead SPI implementation; the CS signal is handled in a way that, in context, makes no sense whatsoever, and could only have been conceived by someone who's never used an SPI interface for anything beyond a stupid-simple classroom example.  Hey, guys?  Some of us use SPI to run long, daisy-chained strings of relay driver chips, effectively being one huge word.  And sometimes we might want to use the MCU in SPI slave mode, with the master shoveling a whole bunch of bytes in and/or out on a single chip select, as one does.  The first case (as well as taking to SPI flash chips) can be worked around by ignoring the hardware CS and using a GPIO pin instead, but insisting on a CS transaction per byte in slave mode makes slave mode utterly useless for most purposes.

Everyday nonlinearity

So, the elderly Prius...

I'd not gotten it to a gas station in quite some time, instead putting in a couple of gallons at a time of lawnmower gas (no-ethanol premium, stored in cans in the barn).

Last week, I'd added right about 2 gallons, if memory serves, and somehow that brought the guess gauge from flashing-bottom-bar to right-around-half.  Which didn't seem right at all.

It was still showing half-ish when I pulled into the gas station, where it took 9 gallons to fill it up, implying that the tank had in fact contained about 3 gallons before filling.  Which is nothing like half the 12-ish gallon capacity.

Anyway, the flashing bottom bar seems to happen at a bit over 1 gallon, I guess, so somewhere a little north of 40 miles to find a gas station.

Someday soon, I need to take the truck and the empty gas cans to the gas station.

A random thought

Am I the only one who can't see "Deglet Noor" without thinking of a hypothetical subgenre of A. A. Milne fanfic?

I think the spiders need a calendar

It's way too early to be decorating their Christmas tree.

Not quite like riding a bicycle

Some things, one can forget.  Remarkably quickly, in fact.  Especially if one is practicing something related-but-different.

Juggling the cars and truck around the past few days, I've often hesitated, trying to remember how to switch between forward and reverse.  Well... the other thing I've been driving that has a steering wheel has a treadle-style go-pedal, so "reverse" is just a matter of pressing on the other end of the pedal.

Switching back and forth with the mower is not confusing, because the controls are entirely unlike those of a car.

Somewhat old news, but...

Can anyone look at a news photo of "Anti" Fascist Jason Charter, and not think: "Now, there's a face that's missing its coal-scuttle helmet"?

Definition flip

I've seen in more than one place the past couple of days the assertion that "Karen" obnoxiously refuses to wear a mask.

Excuse me?  Up through last week, "Karen" was the one obnoxiously demanding that a non-mask-wearer be banished from a store and arrested.  Now, suddenly, she's the one not wearing the mask.

I guess a memo went out from Minitrue.

In this week's episode of Short Attention Span Theater...

Seems the Younger Generation disapproves of the dreaded dot at the end of a sentence.

The explanation given seems to involve SMS messaging, and the notion that any complete thought should fit in a single short message, with no need to break it up into sentences.

Which pretty much gets us into bumper sticker slogan territory.  If it doesn't fit on a bumper sticker, it's immorally complex.

Doomed, you say?

Some Mondays are more Monday than others

Plan for the day: normal start to the day, then grocery pickup at 0800, put groceries away, deal with some paperwork, lunch, then at 1400 take the current feline invalid to the vet for a recheck, then maybe get some stuff done around here.

Well.

First I woke up much later than usual, though not late enough to create a panic, as such.  Breakfast was a quick bowl of granola, though.  Grocery pickup went smoothly, but it turned out that the produce was sparse, in particular things Joy had been kind of counting on for the next few days.  (Also, still no frozen peppers-and-onions blend.)

Then I figured maybe it was past time to make sure the newer Prius and the truck were still OK.  Go find the keys.  Start the truck:  OK.  Leave it running for a while, and go over to the '09 Prius.  Door unlocks.  But... hitting the POWER button promptly drains the auxiliary (12V) battery.  Oops.

There's some rigmarole whereby the rear hatch can be opened with the battery dead, but I don't know what it is.  But wait!  There's a jump point under the hood!  So I open the hood, find the jump point, fetch the jumper cables, drive the truck over, and attempt a jump.

Turns out a Prius with a dead battery draws a lot of current.  Like, more than the truck's alternator wanted to supply.  Maybe if the truck's battery had been fully charged... or if I'd floored the gas pedal... maybe it would have worked.

What actually happened was, the truck's battery drained, in the space of well less than a minute, so low that the engine wouldn't run any more.  Gah!

Well.  Hastily disconnect the jumper cables.  Fetch a charger, a voltmeter, and all the extension cords.  Truck battery is at least at a level where the "smart" charger will recognize it as a 12V battery and actually charge it.  Let it do its thing for half an hour, then start the truck and return it to its proper parking space, then sit there revving the engine for several minutes.  Figure some morning Real Soon Now I should actually drive it somewhere, like to a gas station to top up the tank.  (It also gets driven for recycling runs, next one in probably a week and a half, and gas station runs with a collection of gas'n'Diesel cans, next one maybe in a few days.)

Connect the charger to the Prius.  It works, sort of, for a little while... but, as with the old Prius, there's a rather low threshold where things get nonlinear and the charger gets confused.  Have to disconnect the battery and charge it isolated from the car, and as far as I can tell that requires opening the hatch, so off to the Internets in search of instructions.

Anyway, I think I found the right instructions, so tomorrow morning I should be able to get at the battery, disconnect it from the car, connect the charger, and put some electrons into it, then reconnect it and run the car for a while.  (Idling a Prius in Park does do battery maintenance, unlike, as I'm reminded, idling the truck.)

Oh, and now the aux battery in the '09 Prius will presumably need replacing soon.  At least that's available from the usual sources.  Gotta replace the one in the '02 at some point, and those are harder to find; may end up contacting the local dealership's parts department and asking if they can ship one, because Pestilence.

Addendum:  Maybe I should set up a trio of solar-powered battery tenders at the vehicles' usual parking spots.  Though... connecting to the battery on the '02 Prius gets a bit problematic, maybe, what with the cable probably needing to come out from the trunk rather than from under the hood.  But that car gets driven at least once a week these days, while the other two may sit idle for weeks at a time.

Update, Tuesday morning:  With considerable difficulty, I got the hatch open from the inside, then I disconnected the ground wire from the battery and hooked up a charger.  I'll check in with a voltmeter every half hour or so, and probably switch over to the bigger, "smart" charger once it's over 10V.

The top of the battery does show signs of age, so it was likely about due for replacement anyway.  And, I may have found an on-line source for the '02 Prius battery.  Probably ought to order one of each, to get both cars back into reliable order (though they'll still need to be run regularly).

And, weather permitting, I think after the Prius is dealt with I'll hook the charger up to the truck for a couple of hours or so.

Also!  I looked up the fuse map for the '09 Prius, and, much to my disappointment, there's no 100A-or-whatever fuse that could be pulled, under the hood, to disconnect the 12V battery from everything except the jump terminal.  Methinks Toyota's engineers haven't been maintaining elderly Prii.  (OK, so this problem with the '01-'03 model wouldn't have become apparent in time for the redesign for the '04 model year.)

One more thing I noticed yesterday: the collision-avoidance system in the truck beeps frantically in proximity to tall grass.  I think there's some way to disable it temporarily, if one is going to be driving through, or near, tall grass on a recurring basis.  (We're talking a week's worth of summer growth, here, not full-fledged hayfield.)

Update, Wednesday morning:  The '09 Prius is back in action, though the fuel level is unknown (killing the 12V battery resets the fuel status to "blinking empty"; since the level had been a bit low beforehand, I put in a gallon or two of lawnmower gas), and, this being a Prius, the stereo responded to the power cycle by turning itself on and finding what sounds like an NPR station.  I don't remember if I ever did find an actual way of turning the dagnab stereo off such that it wouldn't turn itself back on at the next startup.  (If memory serves, Inner Party members get an "OFF" switch on their telescreens, but it doesn't actually work.)

I put the Prius back in the regular overflow parking area, and moved the truck to this side of the driveway, where it's now hooked up to the charger.  Didn't want to run an extension cord across the driveway when there are deliveries expected.

Even the simplest things can lack confidence in their abilities

Ever notice how sometimes a gadget will be utterly nonfunctional, but when you walk through all the steps to find out what's wrong with it, each step works, and then when you put it all back together the whole thing works, even though you didn't change anything?

There was one Zerk fitting on the tractor's loader that wasn't accepting grease.  Connect the grease gun, apply pressure, and grease would flow out around the coupler* instead of into the joint and then out around the ends.  Same result with multiple tries, so it's not that I'd misconnected the coupler one time.

I'd already tried poking the little ball in the end, and it wasn't stuck.  So, I figured, the fitting needed replacement.

The replacement fittings (an assortment, since I figure over the years there'll be more'n one in need of replacing) having arrived, I contemplated the geometry of the fitting nestled in its awkward recess, then applied yer basic 7/16" open-end wrench and got it loose.

There wasn't anything obviously wrong with it, so, before replacing it, I figured I might as well make sure it was actually the problem.  Hooked it up to the grease gun, applied grease, and it went right through, just as it ought.  Hm.

Okay, then.  I probed at the hole wherefrom I'd removed it, and 'twas full of grease, with nothing obviously wrong.

Then I re-installed the original fitting, hooked up the grease gun, applied pressure, and the grease came out the ends of the joint, all proper.

Guess it just needed to be shown that it was, in fact, capable of doing its job.

 

* Chrome's dictionary has lost "coupler" now.  It's not exactly an uncommon word.  What's up with the Newspeak Vocabulary Committee over at Minitrue, anyway?

MiniRant: food cans

For some obscure reason, I've been dealing with a lot of canned food lately.  This includes stacking the cans.

Now, a great many years ago, someone had the bright idea of making the bottom of the can slightly smaller than the overall can - a rebated rim, as it were - so that it would nestle securely into the top of the can below it in a stack.

That being the case, how come nearly half the cans I'm trying to stack now are the old pattern, with both top and bottom bigger than the can itself?  Those don't stack worth anything on a pantry shelf... especially if it's a cheap plastic shelf and getting saggy in the middle!  Sure, they can be stacked in pyramids for the Special - this week only! display at the store, but for ordinary storage, it's just a bad design compared to the newer pattern.

And it's not a difference between can sizes, nor between nominal product vendors; we have several cans each of Kroger's store-brand water chestnuts and bamboo shoots, and one product has the stackable cans and the other doesn't.

Yeah, I suppose I could put a sheet of cardboard between layers, if I had a supply of solid (not corrugated) cardboard, and if the stacks I'm trying to make were of suitable geometry for that sort of thing - which, the pantry inventory not being even remotely comparable to that of a grocery store, is not really the case.

Random late-summer Saturday

Been spending some time out at the victory garden, pulling giant weeds that should have been pulled when they were small, harvesting a few tomatoes and squashes, and noting yet more things I got wrong this year.

Now the rain has arrived, so I'm back indoors.

There's yet another set of development tools to be learned, in support of various TI MCU products - I'm particularly looking at the CC26xx series for an upcoming pet project.  Got three copies of the eval board for the Bluetooth Low Energy flavor, but I'm thinking that in practice I'll really want the multi-protocol chip and a somewhat 802.15.4-flavored protocol.  Anyway, I finally got around to installing the TI tools on the new workstation (not yet under the desk, but it's hooked up to port 3 of the new KVM switch, so that's progress).  Now I gotta learn where stuff is.  Anyway, I plan on seeing just what kind of range I can get with the BLE parts, with various antenna options, at various data rates.

Weather tomorrow is supposed to be conducive to outdoors activities pretty much all day, give or take heat and humidity.  Guess I should deal with more indoors stuff while it's wet out there.

Update:  Looks like the SDK for the chip I'm trying to play with doesn't line up with the latest version of Code Composer Studio, and the previous version of CCS has warnings about some things not being supported on this version of Linux.  Haven't actually tried the SDK with the older CCS yet.  Problem seems to relate to CCS and the SDK each bringing its own version of XDC Tools, whatever those are, and CCS can't (for whatever reason) find the one the SDK brought along so it uses its own, which is newer, and some OS function takes a different number of arguments or some such.  And!  The IDE won't let me edit the XDC Tools path to point to the correct version.  Well, enough novelty for one evening; time to be winding down.

A ponderment regarding Chinese infrastructure

Found via Insty, Why Doesn’t China Deploy Fighter Jets to the Spratly Islands?

1. Diplomatic reasons?  (Unlikely, for reasons given.)
2. Aircraft-maintenance reasons?  (Also unlikely, for reasons given.)
3. "The third possible reason, if true, poses a more serious problem for Chinese defense planners: that the structural integrity of the facilities on the artificial islands, including the airstrips, is suboptimal and the PLAAF is therefore wary of using them."

I suggest a variant of (3): they're Potemkin airfields, built to look impressive on photorecon, and the actual infrastructure is nonexistent.  Not saying that it's true (this is pure speculation on my part), but it certainly seems plausible.

Back to pondering smallish machinery

Gotta have some kind of milling machine.  Something, alas, that doesn't require industrial riggers for delivery, nor large amounts of space nor electricity.

So... maybe a Sherline.  Maybe about-the-smallest-thing with an R8 spindle.

Looks kinda like the Grizzly G0781 and the WEN 33013 are fairly similar in specs.  Alas, both have dagnab 16 TPI lead screws, instead of a sensible 10 or 20 TPI.  Wonder if I can find a DRO for either of them.

Also looking at the Grizzly G0758Z, which is a little bigger and heavier and a fair bit more expensive, but it comes with a DRO.  Spindle-to-table distance is maybe a bit short, though.

Ah, well.  Not making a decision this week.  Besides, I still need to build a bench whereto such things may be mounted.

A lathe - bigger than the Taig! - is also very much on the wish list.

For now, though, back to the kitchen.

 

More and different CyberFun!

This time 'twas my turn....

I set out to replace the old VGA KVM switch under my desk with a new HDMI one, while also cleaning the worst of the dust out of the baby server, checking the fans (been hearing some fan noises of late), and adding a WiFi card so's I could stop connecting to the network via a rather limited-speed WiFi-to-Ethernet bridge.

Owing to yet another outbreak of Life, I didn't manage to get the KVM swapout done.  But I did drag out the server, and, er.  Well, I'd meant to add the WiFi card, but what I'd forgotten was that the motherboard has only a single PCIe slot, and that was occupied by the quad Ethernet card I was using to connect to the bridge, with the other three ports intended for the planned house network that still hasn't happened (much pulling of cables required).

So, swap the WiFi card for the quad Ethernet card.  Find that the ath10k driver is installed, but the firmware isn't.  Find the thumb drive with the Debian package for the firmware; install it.  Also comment out the fstab entry for the fireproof USB drive that's under my desk (I'm fiddling with this in the garage).  Reboot.  Bring up the GUI, and sign on to the house network.  Fine.  Fix the Ethernet configuration, as removing the quad card caused the motherboard port to get renamed.  Update all the software for which updates have accumulated in the few weeks since I last dealt with such matters.  Shut down, and return it to its place beneath my desk.

Realize that the WiFi doesn't get signed in until I log in to the GUI.  Look up how to store the password under the NetworkManager hierarchy so it can log in at bootup.  Reboot again.  Boot the workstation.  Discover that packet forwarding isn't working (the old workstation - I haven't switched over to the new one yet - expects to connect to the 'Net through the baby server, since there's that nice gigabit connection between the two machines for file access via NFS).  Find where that is being set up, update the interface names, and reboot again.

Now it's all working as before, including still having the old KVM switch in place (project for a rainy day).  Internet access is slow, but maybe I just need to hook up a better antenna to the new WiFi card (that being why I bought a couple more antennæ of that sort).

Meanwhile, it's about time to fire up the grill.  There's thunder about, but I think the rain is going to miss us again.

Update, Friday:  Pulled out the old KVM switch, the VGA cables, and the DVI cable that had connected the workstation to the monitor.  Installed the new KVM switch.  Had a moment of panic, then realized that it has a power switch.  Now it's working!  Still need to bundle up two more HDMI+USB cable assemblies and run one of them out to the new workstation (to which I'll eventually switch everything over).

Also reconfigured the baby server to use the 5 GHz WiFi instead of 2.4 GHz, since the speed available on the lower band really is limiting.  I don't think that breaks any household networking functions.

I guess I should shop around for a WiFi range extender that truly provides a transparent network bridge and will play nice with one regular antenna and one Yagi or whatever, for bridging the local network to one or more distant stations.

Strange webinar topic

Invitation from NXP:

ACCELERATE SAFE MOBILITY

Um.  I prefer my safe to be immobile and non-accelerating.  Mobile safes have a disconcerting tendency to crash through windows and accelerate at 9.8 m/s² toward unsuspecting pedestrians.  In cartoons, anyway.

CyberFun!

Joy's big project these past few days has been decoupling her laptop from OneDrive.  Because OneDrive is a disaster from just about every perspective, and if you have anything even a little bit confidential on your computer, it's looking like being connected to OneDrive is a Really Bad Idea, even if the important stuff is (as far as you can tell) stored locally.

This means, basically, backing up every-frickin'-thing six ways from Sunday, wiping the old install, and doing a clean re-install of Windows (ugh) 10, then re-installing all the software she actually wants.  Funsies?

Anyway.  This morning, she had everything more or less ready to go, and 'twas time to check e-mail, which meant configuring Thunderbird to talk to her account.  And... it wouldn't log in.  She played with the settings.  I played with the settings.  Thinking it might be the old IPv6 issue raising its head again, I tried replacing the server name with the IPv4 address.  Nope.

(Thunderbird's configuration menu is bloody annoying, but that's a subject for another time.)

I checked the server log.  Many bogus login attempts by some damnspammer looking for a mail relay (sorry, doofus, but only a few accounts are live, and those don't have trivial passwords).  Nothing at all that matches a mail login attempt on her account.

I tried Joy's username and password for an ssh login: username and password OK, but account not configured for shell logins.  Oh, right.  I'd forgotten that.  Well, no matter; she's trying to log in for mail, not shell.

Eventually, she rebooted her laptop and tried the same thing again... and this time it worked.  So, some kind of Windows weirdness.  Surprise!

(Meanwhile, my apprentice at ToasterCo has lately been unable to connect to the Subversion server from work, but can connect from home.  The IT department must have added a new firewall rule.  Or something.  I may eventually have to set up something subversive on the server.)