Archive for May, 2007

Rocket Girls 5

Wednesday, May 30th, 2007

Rocket Girls is cheesily exciting if nothing else. This is exactly why I like Naruto, too.

I half expected to hear, “Aaaaaand the vehicle has clearead the tower!” It never happened, of course. It’s purely an American thing. In Russia, service towers are reclined away from the booster before launch, so there’s no danger of a wind gust drifting the rocket into tower. In Japan, well…

The horribly cheap animation is on display in just about every screencap I took. Sigh.

I don’t quite know why LOX is involved. In real life, supposedly “safe” hybrids use N2O for oxidizer. Perhaps it’s used by OMS on Tanpopo, although usually OMS burn hyperholic propellants. This allows for longer duration missions without a concern for boil-off. Apollo used LOX for something in the SM (I only know this because LOX tank exploded on Apollo 13).

spin_lock bracket

Wednesday, May 30th, 2007

Roland (Dreier, not McGrath) sure likes . Sometimes it produces good results, like when he tilts at windmills powered by . One recent thing he posted was a rant about the proper use of atomic_t. I agree with his point on the whole, but in spinlock case I think he misses something important: spinlocks include optimization barriers, whereas atomic_t only includes CPU barriers (where needed — not on x86). Therefore, in Linux kernel it’s possible to guard a bunch of state with a boolean flag under a spinlock (but not with an atomic).

Some time ago (heck, it was in late 90s, IIRC), did a presentation to a group called “Desktop Performance Council” at Sun, which pretty much was themed “Why Solaris Sucks”. I’m sure it was an entertaining session. At the time he identified the spinlock bracket, such as Roland mentioned, to be detrimental, and pronounced the use of atomic_t superior.

Unfortunately, this was in days of gcc 2.7.2, when we were far more lax about optimization barriers, and since atomic_t did not gain barriers, it cannot be used like spinlocks can, when gcc 4 flexes its muscle. So my recent uses spinlocks exactly the way DaveM lampooned as Solaris’ foolishness 10 years ago.

usblp update

Tuesday, May 29th, 2007

The turned out to be far more annoying than I anticipated. I just sent a patch to test, a month after I wrote “If you can wait a week or so, I’ll get to it.” A week, uh-huh. Ironically, the patch implements Tim’s suggestion about leaving a URB dangling after write(2) returns. I tried a few other approaches, but all for naught. Lessee if this one works.

Of the two other items, spinlocks and dynamic URBs, the first is done, the second is sort of done. I have it all working beautifuly, but the driver grew two new locks. Since one of the objectives was the simplification for maintainability, I am not sure if we can declare success here.

I suppose that there’s no backing out of it now, so, if Tim approves, I’ll forward-port the ENOSPC to merge with the two other items and dump this to Greg or AKPM for shake-out.

Oh, and regarding /dev/usb/lp0e: it’s not happening. The lp.c already has an ioctl defined, LPABORT, so I am going to implement that in usblp.c as a superior solution.

Manabi final is translated

Thursday, May 24th, 2007

suddenly decided to catch up and overtake the awful softsubs.

A glorious day. What are you waiting for?

{Fridge thought: I was in such a rush that I forgot to start screen(1) at mallorn. Heh.}

I blame… Andries Brouwer, I guess.

Thursday, May 24th, 2007

[root@niphredil zaitcev]# dosfslabel /dev/uba2
IPOD
[root@niphredil zaitcev]# mount -L IPOD /mnt/ipod
mount: no such partition found
[root@niphredil zaitcev]# 

A mystery. This works fine for USB keys, but not iPod.

Lucky Star References

Thursday, May 24th, 2007

One thing I find strange is how old some references are. The Attack No.1 was in early 70s, if I remember well. Not 90s, not 80s. I don’t know when the Timotei ad was broadcast, but that does not sound very recent either. How in the world can Konata know about these things?

I blame Konata’s father. We’re talking about a guy who knows how to tie a sash for crying out loud. The ep.7 afforded us the first glance of him. That high nose was something else. That guy must have some stash of ancient cultural artifacts around the house. Maybe a real set of swords from before WWII. Or VHS tapes from the 80s.

Web Passwords

Thursday, May 24th, 2007

Thomas to share the password management scheme.

I used to use a limited set of passwords, but found that solution inadequate for a couple of reasons. Firstly, sites to which I do not want use them proliferate, e.g. my mortgage company, my stock trading accounts, my banks, my cellphone accounts. They need their own solution anyway. Secondly, I can never remember which password was used where. Once the set grows bigger than 3, one starts running the risk of being locked out when trying.

Eventually I gave up, wrote them all into a text file, and encrypted the file with gpg. This seems to work well.

I also allow Firefox to remember passwords for forums, Mixi, etc. I do not know how strong the Firefox’ encryption is with the master password, so I do not trust it with sensitive passwords.

My biggest issue currently is Google and its Borg bend. Their single sing-on scheme covers very sensitive stuff, like Adsense. In the same time, I have no idea how safe Gtalk’s authentication is against snooping. Even it Jabber itself is strong, Pidgin as an implementation, frankly, does not inspire a lot of confidence. I mentioned before that it does not seem to have any support for a key store. Clearly it wasn’t written by people who care much about security.

Cheating

Wednesday, May 23rd, 2007

I’ve got Ranma Season 1 on a thinpack today for my fansubber’s duty.

The reason this can be considered cheating is that I watched Ranma about three years ago, when thinpacks weren’t available. Thus, by procrastinating sufficiently, I am only paying VIZ about $40.

I probably won’t even open it. OTOH, this is what I thought about Chobits too…

BTW, one more thing. The copyright is from 2001 on this release. The original broadcast was in 1989. So, basically, studios are already getting the Lessig’s perpetual copyright, only without paying any duties. I wish homeowners were so slick.

Xenoglossia is really freaking creepy

Sunday, May 20th, 2007

So Haruka is in love with the giant robot she’s piloting, and the feeling is mutual? I remember how reviewers loved to write about Freudian significance of getting inside the Eva unit made by your father. This shit is right there with that. When the 2006 brought us giant mecha powered by lesbian power, it was not quite as creepy, because of more open and unabashed nature of it (also allowed to reject the show sight unseen, which was useful). But this is sneaky.

Otherwise, Xenoglossia seems like a neat enough show. Pretty animation; big budget brings collar bones back, hip-hip hurray. I wish they didn’t try to extend it for 26 episodes, but it’s hopeless. There doesn’t seem to be enough character story, but there’s enough of the detective story.

As an aside, this is funny:

They have actually said “kurake(ru)”, so the caption is not a fantasy of the translator. The ESR’s crusade starts to make headway just when ESR himself decided to marginalize himself. Ironic.

Back from FreedomHEC

Saturday, May 19th, 2007

I’m back from . In theory, it’s supposed to be a place of advocacy and tender loving care for representatives of hardware vendors. I came there primarily to answer questions in hallways, such as they are. In reality, it turned out to be a flash meeting of key Linux driver hackers: Greg Kroah for device model and leadership, James Bottomley for SCSI etc., Jon Masters for out-of-tree and modutils, for udev, and finally David Zeuten for the borough of HAL/GUI. I substituted for Dmitry, I suppose, although obviously I mostly watched them mulling things. Stephen Hemminger represented the network drivers. I did not communicate with a single vendor’s engineer who weren’t already involved with Linux.

Between sessions, Kay showed me a prototype of enhanced udev in combination with a “small” patch, which allows to abandon libusual. This is a huge deal for me. By itself, ub is an excellent driver, but I was really desperate when I came up with libusual, and, not to put too fine point on it, libusual is garbage. But since it is (or was, now) necessary for ub and usb-storage to coexist, libusual pulls ub down. In context of bug , I recommended for Fedora to stop shipping ub. Fedora was the only distro which ever tried shipping ub (Unless you count Gentoo, where users pretty much configure what they want, as I understand.). And now ub is saved by a serendipitous coincidence.

I might want to keep libusual in order to gain access to useful flags (such as off-by-one capaicty), but that is acceptable.

James gave an outstanding talk about git. Such talks usually come out as regurgitation of trivial things which I know already, but this was different. I knew what he was talking about, but he turned things so I was able to understand them. For example: “… allows multiply heads, called branches” — So this is what branches are! So simple and obvious!

Another thing which he did (probably unintentionally) was to show that I can ignore the scary message “It looks like you’re committing a MERGE” in commit comments. He exited with “:wq” and that was all. I always quit at this point, and that left the tree in the state where magic and dangerous manipulation of git-update-index was necessary to make any progress.

He also ran down git-rebase, which I blissfuly ignored.

I learned a few other interesting things. ARM comes out with an SMP (called “Cortex”). Debian supports out-of-tree modules with something called “module-assist” (does it load nVidia too?). Bums in L.A. are washed and dressed well; also they are extremely persistent. Hitachi spun off SuperH into “Renesas” (W.T.H. with all these names?). Corolla is actually a pretty crappy car, despite being the most popular car in the world (the only good thing in it is the auto transaxle; why could not Toyota use the same tranny map in RAV4?).