Hypothesis: Sarah Palin resigned to accept a secret appointment as Secretary for Extraterrestrial Relations. After all, she can see Tau Ceti from her house.

Well, at least my theory is more entertaining than the mainstream media....

Wow! VNV Nation are fucking *amazing* live.

I've griped about Oddfellow's not having a menu on their web site before, but ... wow. This is most emphatically not an improvement. Standards-free Flash crawling horror, complete with mystery meat navigation and enough bloat to choke a tyrannosaur on.

It's like something out of a Lovecraft story. "The most merciful thing in the world, I think, is the inability of open software to correlate all its contents. We live on a placid island of RFC compliance in the midst of black seas of binary horror, and it was not meant that we should voyage far. The open source flash projects, each straining in its own direction, have hitherto harmed us little; but some day the piecing together of dissociated data will open up such terrifying vistas of reality, and of our frightful position therein, that we shall either go mad from the revelation or flee from the light into the peace and safety of a VT-420."

This would be why Flash is banned from my personal machine.

Wow, The Stranger published something I actually really like: Shocked and Repelled: Our Dangerous, Depraved, Sometimes Hilarious Sex Lives and especially this article. When repressed, conservative straight people say "You people are all just a bunch of deviated preverts!", the correct response is not "I'm just like you, though. My sex life is a half-hearted reenactment of culturally prescribed gender roles, marred by gnawing insecurity about whether it's really normal to fantasize about her spanking me like that, but it doesn't matter because I'd never in a million years have the guts to actually ask for it anyway. I want to live in the suburbs, have 2.3 kids and watch television every night, I swear it!" The correct response is, "Yes, and it's adventurous and exhilarating and absolutely wonderful, and we both know that you're really just jealous, but you're too afraid of yourself to ever admit it."

Disclaimer: I don't think there's anything wrong with you if you actually want to live in the suburbs and have 2.3 kids, as long as you're not judging others for wanting something different. I do think an awful lot of people just follow cultural default patterns without really thinking about it, and that's kinda sad. On the other hand, if you're one of those repressed conservative people who want cultural norms or even enforceable laws that impose that kind of life on others so that you can avoid dealing with your own insecurities, yes, there is something sick and wrong about you, something rotten at your very core.

Doctors Baffled, Intrigued by Girl Who Doesn't Age - ABC News

Huh. Maybe there's some hope for SENS after all. Then again, maybe this is just a severely belated April Fool's Day joke. I did first see it on Slashdot, after all.

The two replacement drives I order arrived today; now everything seems to be working. I'm rsyncing stuff over now. *crosses fingers*

I'll be really happy when mass storage is all solid-state someday.

I gave up on trying to get the new disk array to work as a single > 2 TB volume and used it in split mode with LVM instead, and then while I was copying stuff over to it a couple days ago, it started throwing I/O errors. I tried unmounting and power-cycling it, and one of the drives I had re-used from my original enclosure wasn't spinning up, and the other one took noticeably longer than the two new ones. Fortunately, I still have copies of everything elsewhere. I guess running for a while on screwed-up input voltages must not have been good for those disks. Now I don't trust even the one that does spin up, so I ordered another pair of matching disks, and tomorrow I get to try all this again. That really should be the end of it, seeing as at that point it'll be a brand new enclosure *and* four brand new disks. This is getting annoyingly expensive. Now I need to figure out what to do with a 1 TB SATA disk that seems to work but that I don't trust at all now. I don't think I'm quite unscrupulous enough to sell it to some unsuspecting person.

This required quoting for posterity:

Andrea: see, people always worry about losing weight, but never about their higher multipole moments
Dana: hehe
Andrea: i say it's time for more spherical people!
Dana: yes!
Andrea: we'll, um ... waste less energy on gravitational radiation if we don't have quadrupole moments!
Andrea: you don't want to waste energy, do you?
Dana: hehe
Dana: :)
Dana: can't waste energy :)
Dana: it's conserved

I'm pretty sure I've ranted about this before, but why is it so bloody difficult for web forms that take credit card numbers to just strip dashes and whitespace? I'm trying to pay my T1 bill and Speakeasy's form insists on the credit card number without dashes; if you enter it with them, it says 'authorization denied'. Presumably, it's passing the number to something upstream which is strict about it. It then bounces me back to the original form, where I have to type my name and billing address all over again. Weirdly, the state of my billing address and the expiration date of my credit card are preserved, but nothing else, probably because those are drop-down menus and some UI designer somewhere just went with the default behavior without bothering to engage his brain for five seconds.

It's perfectly reasonable for an API exposed by a credit card processor to a web server to require strict formatting, but this sort of rigor should not be exposed to the user. Credit card numbers are commonly written with dashes, lots of other web forms will reject them without the dashes, and I shouldn't have to remember which particular rules this form insists on every time I use it.

Why didn't anyone tell me about Paleo-Future before?

Evening fashions of the year 1952 - as predicted in 1883

"The American will be taller by from one to two inches ... He will live fifty years instead of thirty-five as at present - for he will reside in the suburbs. - Ladies' Home Journal, December 1900

The READ CAPACITY 16 thing didn't work and asking in linux-scsi produced no answers at all. I just caved in and used the split mode and LVM instead.

Woohoo! We're all going to DIE!

Holy crap, Dick Cheney supports gay marriage.

*looks out the window for a rider on a pale horse*

It would seem that the queer-hating fundies now deign to associate with us Capitol Hill sinners. I am overwhelmed with glee.

I'm rather tempted to show up at Grey on Monday to see if anything amusing comes of this. It'd be more fun if Dana weren't so far away, though. Staging a very conspicuous lesbian make-out session at the next table over seems like about the right response to this. :)

The READ CAPACITY 16 thing doesn't seem to work; the command returns a SCSI error. Now I'm really curious what's going on. This disk array has a mode switch on the back, like I mentioned in my previous post, so presumably it has *some* behavior that allows access beyond 2 TB, but this mechanism is not the READ CAPACITY 16/READ 16/WRITE 16 normally used for large SCSI devices. The manual for the array seems to imply that the 64-bit versions of Windows can deal with this, but I don't know how they do it, and without a hardware USB protocol analyzer it seems unlikely that testing that would yield any insight beyond knowing whether or not it does work.

The people over at usb-storage don't seem to know either, unfortunately. The suggestion that Windows just doesn't use READ CAPACITY at all and goes by the size in the partition table entry seems unlikely to me, since it would need to know the underlying disk capacity when creating the partition table, and since it would also need to use READ 16/WRITE 16 to access past 2 TB even if it knew the capacity through some other means. I suppose I should try asking linux-scsi.

The present weather in Seattle makes me want to move to Pluto and live in a solid nitrogen igloo.

It turns out this a confluence of what seems plausibly like a hardware bug and some workarounds for other hardware's bugs. The old SCSI READ CAPACITY command only has a 32-bit field for returning the maximum block address, so for devices larger than 2T (assuming a 512-byte block size), something else is needed: READ CAPACITY 16, which allows a 64-bit block address. The problem is that, apparently, some devices hang when given READ CAPACITY 16, even some recent ones. Further, USB devices (the USB mass storage class is a wrapper around SCSI, essentially) are all assumed to be SCSI-2, and READ CAPACITY 16 is apparently only in SCSI-3, so any support for it is unofficial and of necessity hacky.

The upshot of all this is that, when probing SCSI disks, Linux tries the regular read capacity command first, and then tries READ CAPACITY 16 only if READ CAPACITY returns 0xffffffff. Presumably, devices larger than 2T are supposed to say that to signal the need for READ CAPACITY 16, and to allow older systems to see as much of the device as possible without it. This disk instead returns the low 32 bits of its actual size, inexplicably plus 12, hence my suspicion of a hardware bug.

Anyway, to fix it, I would need to force READ CAPACITY 16, which presumably would produce the correct value, but in such a way that we don't use it on other disks that it might cause problems for. The really right choice would seem to be a quirks table, with a flag on the scsi_disk structure based on the vendor/product strings to indicate to always try READ CAPACITY 16 first.

My disk array has a compatibility switch for systems which don't support devices larger than 2T, so I can use that to force it to present as two smaller disks and then run LVM over them. That seems to work, but if I go down that path I'll never get around to fixing this kernel issue because I won't want to disrupt my filesystem once I'm relying on it. I'm trying to decide whether I feel up to implementing a quirks table, or to just go with the LVM option.

Now I have my new disk array set up, and plugging in the USB cable yields this:

usb 3-3.6: New USB device found, idVendor=0dc4, idProduct=0004
usb 3-3.6: New USB device strings: Mfr=1, Product=2, SerialNumber=3
usb 3-3.6: Product: MGB Quadro
usb 3-3.6: Manufacturer: GalaxyMetalGear
usb 3-3.6: SerialNumber: 540323
scsi 4:0:0:0: Direct-Access     External  Disk 0          1.15 PQ: 0 ANSI: 4
sd 4:0:0:0: [sdb] 3519133184 512-byte hardware sectors (1801796 MB)
sd 4:0:0:0: [sdb] Write Protect is off
sd 4:0:0:0: [sdb] Mode Sense: 10 00 00 00
sd 4:0:0:0: [sdb] Assuming drive cache: write through

This, rather suspiciously, is almost exactly 2^32 sectors off (it's high by 12 sectors) from the correct count, 7814100672. I sense a 32-bit field somewhere. This might have something to do with it.

Update on the disk situation:

The LaCie support tech got back to me with a standard-issue "We only support Windows and Mac, you would need to buy a whole new machine you don't have and don't want just to prove to us that it's broken" spiel. Yeah, because it's not like the USB mass storage interface is a standard or anything like that, or like software on the host couldn't possibly cause a disk to not spin up before the cable is even plugged in, or couldn't make the power brick emit audible hissing sounds. *sigh* Well, I'm glad I went ahead and ordered a new enclosure anyway.

Also, I'm awarding bonus fail points for sending support ticket update e-mails in HTML-only.

My 2 TB LaCie Big Disk Extreme suddenly stopped spinning up last week. Noticing the ominous promise to destroy my data if I sent it in for repairs under warranty on their web site [1], I opened it up found a pair of Seagate 1 TB 7200 RPM SATA drives connected to a small controller board. I managed to get it running once more by disconnecting the power connectors to the drives, and then reconnecting them one at a time to spin up the drives consecutively, so I copied most of its contents to my laptop, and pondered what to do from there. I ended up ordering a 4-bay enclosure that seems much more solidly constructed than the other one, and another pair of matching drives, and then a portable 160 GB drive to hold the rest of my data temporarily that I hadn't had space for before.

Anyway, all those parts arrived yesterday, so I spent most of Friday evening copying things around. Mostly pretty routine, but getting the old enclosure to work one last time to copy my remaining 120 GB of data was ... Interesting. After having sat idle for a week, it seemed less functional than ever. Now, the power brick, a switching power supply with +12 V and +5 V outputs fed directly into the drives and also powering the controller board in the enclosure, was making a hissing sound audible from up to about 50 cm away. I can usually only hear power supplies at very close range, so presumably it was operating at a much lower frequency than intended and hence more easily audible. This time, no amount of playing with the timing could get either drive to spin up at all. Anyway, I figured my last chance was to power the drives from a different power supply. I was a bit worried about having the drives and the controller running from potentially different grounds, especially if the problem with the original power brick had left the output floating or something, but it was all I could think of to do, so I set about performing surgery on a Molex Y-cable to connect it to the SATA power connectors, and running it from a spare Sun Ultra 10 I had around.

It worked perfectly and I rescued all my remaining data, and now I can reuse those drives in a nice 3.8 TB RAID-0 setup. Now, this isn't quite as impressive as the time I got RS-232 to run at 9600 bps with full flow control over a few pieces of aluminum foil, but I'm rather proud of it, seeing as I rarely touch hardware at all these days.

[1] See threat of data oblivion here. Why would anyone even consider doing that? The problem is much more likely to be the controller board or power supply than the drives themselves, and I'd much rather spend $400 on a new drive and rescue my data from the failing one myself than get a new one free and lose all my data.