Silva Tulga's Journal|
[Most Recent Entries]
Below are the 20 most recent journal entries recorded in
Silva Tulga's LiveJournal:
[ << Previous 20 ]
[ << Previous 20 ]
|Wednesday, April 1st, 2009|
Riding log for Tue Mar 31
Ljufur wasn't expecting me to actually want to bring him up, but was glad when that turned out to be what I was there for. When I got him up to the ring, I walked over to the scary corner, where he looked around a whole lot trying to figure out what was so scary about that corner. I chased him around a bit, and then got out the bouncy ball and chased him around some more. He was quite eager to run, going between a trot and what might have been a flying pace when I had the ball out. He didn't want to run along the fence toward the barn, and kept turning around in the corner by the flagpole instead, but would go to the left fine in the rest of the ring.
Then I groomed briefly and got on. He didn't particularly take off when I got on, but he sort of wandered forward. However, he was kind of stuck in the corner, so he didn't go anywhere, and when I turned him away from the corner and told him to stop, he stopped fine.
We walked around a few times in each direction (since I wasn't really warmed up), and then tried the tolt. He kept pacing instead until I tried going up to the canter for a stride or two and then back down to the tolt, which came out very nice, and he went around pretty much until I said something about it. I haven't been able to convince him that there's something that means "good, keep doing that", rather than "good, now enough".
Did some stretches and a bit more brushing, and brought him down.
|Monday, March 16th, 2009|
Riding log for Sat Mar 14, 2009
Brought the boys up to the ring and chased them around while V. set
up a jump.
Groomed for a really long time; lots of shedding, and he'd gotten
muddy for the first time in a while.
Put on his saddle; small girth, 3rd and 4th holes(?)
Had a hard time getting the bridle assembled correctly; might want a
symmetric bit to avoid having it matter if it flops around in the
floppy bridle. Having the reins right might help, or might make it
Practiced getting on with a saddle (and stirrups) a whole bunch. Doing
it like vaulting doesn't work with the pressure on one side. Might
want to ignore the stirrup and vault on. Ljufur kept taking off as
soon as I was in position, until I got it right, at which point he may
have also just decided that he would have to stand there and be bored
until I asked him to go, or I'd just get back off.
When we started walking toward the gate, he felt goey, and he'd been
running away with people on the trail, so I said I'd like to run
around in the ring a little first, especially because we were just
going to walk while on the trail.
Got a pretty nice tolt after a bit of asking (mostly just by doing it
a bunch, rather than doing anything specific). Didn't want to tolt on
the remaining snow, and we had to get room between the jump and the
snow before he'd go around without breaking. Did pretty well at the
tightish turn at the narrower end of the non-snowy part of the
ring. Not sure if he tended to keep tolting through it, but was
definitely tolting coming out of it.
Went on the trail; heading for the gate, we were ahead due to where
we'd been when we decided we were ready to go, and had to stop to let
Stjarni go ahead. Ljufur seemed fine with the idea of going to the
scary corner and out in front, though.
No hesitation going into the water jump right after Stjarni, but kept
his head right by the water; don't know what he was looking for, since
I don't think there was any way to see anything useful through the
water. He was still low and forward when we got to the end and stepped
Tried to keep his nose in Stjarni's tail through S's call of
nature. Silly pony.
No hesitation crossing the street. Clearly used to it by now, or
Stjarni being ahead made it okay.
Went up the big hill with no problem; I kept remembering the first
time he'd been there, and had treated the steepish parts like there
was a wall there. I still think it's kind of precipitous coming
down. On the actual trail, we didn't have anywhere he could really
run, since we just went up and turned around and came down, but he
didn't show any sign of wanting to run away with me.
On the way back, after the street, he kept falling behind Stjarni's
big walk, and jogging slightly to catch up, but didn't feel like he
wanted to pass or race or anything. Stayed a bit closer to Stjarni
than I'd have liked in retrospect crossing the puddles in the path,
since that gave us less stumble-room.
After getting back, we closed the gate. It took a bunch of tries to
get in position, in part because there's still snow along the rail,
and he was avoiding it by default, but that didn't put us close enough
for me to reach the gate. Once I'd gotten the gate away from the
fence, he wanted to help, so I backed him out of the way to where he
could push it past him. Had to go around a few more times in order to
be able to reach it to pull the latch, and he stood nicely despite me
leaning funny to reach.
I needed to stretch my ankles before I could do half-seat at all. Then
it was okay for a bit, but my legs gave out before too long.
We were trying to trot along the snowy side of the ring, following
Stjarni (so he'd know what we were trying to do), but he actually
wanted to pass Stjarni the first time, and the second time V. took
off while while we were trying to get behind her, and then Ljufur
cantered a bunch of catch up and then paced, and my legs were tired.
V. tried again, and he offered a very nice 4-beat tolt, and also a
very 2-beat pace, but didn't trot at all.
Untacking, Ljufur stood very nicely for me as I was taking his bridle
off and putting his halter on, despite V. bringing Stjarni down in
the middle of it and leaving the gate open at the top.
|Tuesday, January 20th, 2009|
Google Docs is really cool. They actually managed to come up with a better way to do crosswords with a bunch of people than sitting around a piece of paper with a pencil, because every participant can write in the grid at the same time without a tangle of arms blocking everyone's view. There were puzzles we did this weekend with four people sitting at a table working on it discussing suggestions, with four laptops open to the same spreadsheet. We also did puzzles with more people working on the same document at the same time than the interface has distinct colors. And I did a puzzle from home with someone somewhere else, and I don't know who it was or where they were.
I think the most annoying flaw with Google Docs is that if you have 100 documents and 20 people, you have to set up 2000 sharing relationships, rather than being able to put all of the documents in a folder and share the folder, and being able to define a group of people and share with the group. Ideally, you'd be able to say, "these people are my team; these documents go in this folder; share this folder with that group". We mostly got around this by having a single account that most people used, which works remarkably well, considering (I wonder if Google noticed that one user was occasionally making changes to the same document from a dozen computers, probably in 4 time zones). On the other hand, it was annoying to have no identifying information about the other people on a puzzle.
Of course, being the decentralization wonk I am, I now want to set up a Google Docs without the central server. I think it should work by having people with XMPP (Jabber, Google Talk) clients exchanging updates end-to-end (although maybe mediated through a referee, probably whoever created the document and shared it, although it could be delegated to a well-connected entity). The right thing is probably to use ODF as the format for the document, and define a set of edit operations that can be made to XML documents that are reasonably localized, as far as is possible, and then have people pass each other these edits.
|Tuesday, October 21st, 2008|
Evidently, suspend to RAM is the new audio in Linux. It always used to be that, when I tried to get audio working in Linux on a computer, it would be completely impossible the first time I tried, so I'd give up, and then I'd try again a while later and find that it was working perfectly already. A while ago, I tried to get suspend-to-RAM working on my laptop, and had a lot of problems. I just tried it, and it works perfectly. It even works twice in a row. I didn't even have to undo any of my previous attempts to configure stuff. Even controlling it with the lid switch works (although I did have to put something in to do that).
I wonder if my modem works.
|Monday, October 13th, 2008|
I've been reading Zogby's polls, largely because they seem not to be too confidant in any particular result. But I may stop reading their press releases after today's
, because it looks like they've never taken statistics. (This is a bit surprising for me to notice, since I've actually never taken statistics, being in a field where, generally, if something ever happens, that's really interesting.)
They make a big deal of the fact that their polls show Obama slipping 1% and McCain gaining 0.8% in the poll data taken on Sunday; but a quick look at the table shows that this almost but not quite compensates for a swing in the opposite direction in the data taken on Saturday. Furthermore, they give the margin of error as 2.9%. I'm pretty sure that, with the assumption that the true values are continuous functions dominated by low frequency components 1/(2 days) and below, that the most likely explanation is that the 10-11 data was a statistical anomaly. In fact, I think that, given the margin of error, we should expect an anomaly of that size about that often For that matter, nothing particularly notable happened Saturday or Sunday that would logically incline voters to change their minds on those days, so far as I can tell.
For that matter, I suspect that their 3 day window would give odd window effects if anything particularly interesting actually happened on a short time scale; there seems to be as much weight given to the polls from 3 days ago timing out as the newly introduced data. Based on this, I predict that McCain will get about a full point from Obama in the 10-14 results, which will disappear the next day. And I suspect that Zogby will make a big deal of this, despite it actually meaning nothing at all other than they don't know how to filter noisy time-varying signals.
|Sunday, May 25th, 2008|
I'm reading Livejournal at the moment primarily to evaluate different methods of formatting comments, so I can decide what defaults I want to use for my web app. It's kind of odd to be looking at things I normally exclusively for content for form instead and noticing how much variability there is that I hadn't noticed.
|Thursday, May 15th, 2008|
|Friday, February 1st, 2008|
Among other things, today I fixed a 1-character bug that took 19 months to find. My calendar web page elements weren't having the columns set to be equally sized while that size would be selected based on the contents (either I'd fix the size of everything, or different days of the week would be different sizes), because browsers were pretty much ignoring my col-group tag. When I finally gave these web pages to a validating XML parser with the XHTML DTD that they reference, it pointed out that that's supposed to be "colgroup", and with that corrected, it all works like I'd always thought it should.
|Sunday, September 23rd, 2007|
For the first time I know of, I've been photographed
by a Linux kernel maintainer. Of course, I'm impossible to identify in the picture, because I'm wearing dark sunglasses. I appear in his picture, and he makes my laptop battery not run out so fast.
I'm disguised as a total stranger. In particular, this person
. If I'd ever seen her, I'd have worn a hat to be more convincing.
|Tuesday, May 15th, 2007|
It's great when somebody asks for a feature that sounds simple, but actually takes a couple of weeks of work to get right, and you respond four hours later (having not read the email for most of that time) with a patch that does it. Conveniently, I'd gotten the couple weeks of work done a couple of weeks ago, and gotten most of the tricky stuff in (queued for the next version) over the weekend.
|Monday, April 30th, 2007|
I suspect Linus Torvalds is going to write a really neat bug tracking system in a couple of weeks. I wonder if I want to work on it. I wonder if I'd be able to make a credible program first by trying to write what Linus would write. The really weird thing is that my mental model of Linus produces a totally different design than what I'd personally write as myself, and, weirder still, it's starting to convince me that it's better.
|Thursday, April 5th, 2007|
I just found Volity
, and was curious as to whether it would be difficult to implement 5-suit pinochle for it. Then I was curious as to whether I still had the rules. Turns out that I do:
The deck consists of 61 cards: 5 suits (stars, spades, clubs, diamonds, and hearts), 6 ranks of each (9, jack, queen, king, 10, and ace), 2 of each of these, plus a single joker.
The teams are as in standard pinochle. The deal is as usual, except that each player gets 14 cards and there are 5 in the blind. Bidding starts as if the dealer had bid 24. The bidding and calling trump is an usual (except that trump can be stars, of course).
For purposes of meld, the person with the joker can declare it to be any single card.
Something around requires one card of each of the 5 suits, but one card of each of 4 suits is worth half the points. Only the highest-scoring combination of a rank counts; if you have 9 jacks, you get 20 points (double incomplete jacks around).
In addition to the usual pinochle, a king of hearts and jack of stars is worth 4 points, or 30 for double. One of each kind of pinochle is only worth 8 points.
The play is as usual, except for the joker. The joker never wins a trick. If the joker is led, the next player can follow with any card, and that card's suit becomes the suit to follow. The joker does not count as following suit.
One option involving the joker is to allow the joker to count as throwing high trump for the purposes of following suit (that is, if you are out of the suit led, you can throw the joker, if you want, instead of a trump, even if trump has already been thrown; but you can't throw it if trumps were led and you aren't out of trump).
Another option involving the joker is to make it a point card.
If the joker is not a point card, there are 31 points available in the play; if it is, there are 32.
The rules for joker pinochle can be applied without the fifth suit, but 5-suit without the joker doesn't work too well, IIRC.
|Sunday, March 18th, 2007|
I'm starting to get quite a collection of protocol violations in various jabber programs. E.g., livejournal.com seems to pass on requests directed to a node to a connected client (it's supposed to answer with different information on its own), and Adium (actually, libgaim, most likely) tries using "last" and "version" without first checking whether they are supported.
I think it's a bit odd that neither livejournal nor google seem to have their own multi-user chat, but they do both provide access to services elsewhere. Actually, I'm not sure if there's any way to find out whether a service has associated multi-user chat, because it has to be a different domain and there's no clear way to report what that is.
Still have to get back to doing the tricky parts of having a bunch of clients using XMPP over a unix socket to divide the client-side functionality. Oddly, that's mostly figuring out how I want to pass requests through unmodified, or whether I want to intercept all of them. But the gtk client seems to be working, anyway.
|Tuesday, March 13th, 2007|
This Sunday, I tested for my black belt in aikido. This consisted of four sections: being stable while people pushed me, demonstrating a variety of techniques (my choice) against a variety of attacks (specified); kata with (wooden) sword; and freestyle. I'm writing this from memory, but there's a video I haven't seen yet, so I'll probably update this later when I see the video.( Read more...Collapse )
|Monday, February 19th, 2007|
I bet it would be an interesting addition to my email handling system to automatically sort messages based on various criteria, eventually coming down to date, whenever they've been in my inbox for a month. Then I wouldn't end up having to move 38191 files at the same time.
|Sunday, February 11th, 2007|
It would be interesting (and probably incredibly useful) to have a weather forecasting site that represented the actual extent of the forecasters' information about the weather, rather than just best estimates of certain values. Then it would be an interesting UI problem to allow people to query it in such a way as to get the best results.
For example, there's currently a reasonable likelihood of a nor'easter off of Cape Cod on Wednesday. But that's just their best guess as of this morning. On the other hand, this afternoon, they should have measurements which will make this evening's weather report a much better guess than this morning's, by more than the amount that 3-day-out forecasts normally improve. So their advice this morning would probably be: don't count on it not snowing on Wednesday; if possible, check the evening's forecast before rearranging your plans, though, because tonight's when we'll go from speculation to a more sure idea.
There's also a lot of information about forecasts that they know pretty well, but can't actually report in the usual format. E.g., "It'll snow lots this weekend, but we don't know where. It'll be somewhere between Boston and NYC. If you're in any of these places, there's a small chance of snow. If you drive from Boston to NYC, you'll be sad." Or, "Big storm, some time between Tuesday and Thursday, otherwise clear. Don't expect to be snowed on any particular time, but you'll definitely need your snowshoes if you want to walk in the woods on Friday."
|Tuesday, January 16th, 2007|
I seem to have lost my mind
Missing: my mind. Description: it contains an idea of the form "Despair.com should make (some device or item normally involving a positive noun)
". Last seen: my second-floor kitchen.
If found, please return to this journal.
I have the feeling that it'll turn up in somebody's list of quotations, correctly attributed to me, and I'll have entirely forgotten ever saying it. This is the sort of thing that happens to me.
|Monday, January 15th, 2007|
Supporting fseek on a network connection is a bit tricky, but not as bad as you'd think if it's HTTP. Turns out that I remember how to use IRC. (Except I think the firewall doesn't handle DCC.) Also, svn commits aren't too difficult. And shoutcast is pretty easy, too, at least if you're not using UDP.
So, now I'm officially an audacious developer. I'm not sure if that means I'll be changing all the plugins to make performance not suck when there's significant latency to opening a file repeatedly.
Also, the reason I'm sort of sleepy is that it's 3:30 again. That keeps happening.
|Tuesday, December 19th, 2006|
Finally got sufficiently fed up with audacious pausing for a while after every track that magnatune.com streams to do something about it. In particular, I added a line to the middle of a string constant. Really, I ought to rip out a bunch of code and replace it with calls to a library, or at least rewrite the loop code to notice the end of the logical stream, but the string constant works perfectly, so...
The part I added was "Connection: close\r\n". Obviously.
Of course, I probably want to do it more right later, because I'd like the code to recover if you pause for a while and the connection times out, which means identifying that you haven't reached the end, and reconnecting with the range set to where you were.
|Tuesday, December 12th, 2006|
From the linux kernel 126.96.36.199
Daniel Barkalow (1):
forcedeth: Disable INTx when enabling MSI in forcedeth
Of course, the mainline (2.6.20 and beyond) patch for the same problem is different and by somebody else (and solves it for other hardware as well).