D70 Trap Focus Trick

August 19th, 2004 | 11:02

From this DPReview.com post, you can set up trap focus as follows:

  • Camera should be in autofocus single-area
  • “Pencil” menu items should be set so that 03 AF-AREA is “Single” and 15 AE-L/AF-L is “AF”

With these settings, the shutter release button doesn’t trigger autofocus anymore: item 15 tells the camera that the AE-L/AF-L button should trigger autofocus.

Once this is set, and if you or the focusing target move so that it’s out of focus, you can hold the shutter release all the way down, and continue to hold it. When something is in focus, the camera will fire off (in a test, you can back away from the subject after hitting the AE-L/AF-L button to get focus, and then come back in while holding the shutter release button down). This trap focus technique is typically used by, say, sports photographers who set focus on home plate, and then wait for the runner to slide home: the camera reacts faster than the human.

Neat trick: I had no idea the camera could do this.

More Judo

August 16th, 2004 | 16:49

This past Saturday was my fifth or sixth judo class at Oishi’s, but I got to play with a yellow belt on the ground (instead of the usual situation of having a black belt sit on my chest and give helpful hints on how to get out of the hold that work if he’s not paying attention). Interestingly, we were fairly evenly matched. In the first round, I got a nice choke in (a jigoku jime variation, I think) after a few minutes of working against his turtle. In some subsequent round, I almost got a hammer lock on him when he tapped. In another one, from our suwari-waza starting positions, he reached for my lapel, and I did a variation of the “bring the elbow out of his center and cut down his forearm like a bokken” technique (I’m sure there’s a nice name for it in aikido land. Update: it’s one of the kokyunages; the closest on the referenced site is katatetori kokyunage, but directly with the irimi and without the tenkan), which brought him down to the ground quickly and let me open that phase by being on top. I think I had a holy-crap-that-worked look for a while.

I’m, of course, leaving out the various times he got me. He was also nice enough to show me another way to attack the turtle position, and we went over the kataha jime choke. I still don’t know how to work from or against side control.

I also got one of the brown belts in a variation of yoko otoshi on the first round of randori, when he was apparently thinking, oh, a white belt. After that, I didn’t get close (except when he told me to do yoko otoshi again): he kept on foot sweeping me and threw me a few times in tani otoshi (I don’t think we’ve done tani otoshi in Eizan Ryu, mainly because tori and uke are fairly entangled on the ground, and we prefer not to get into that position). In an earlier class, I worked with a yellow belt in randori with similar results but more parity. He did fewer sweeps (possibly because his legs were relatively short) but instead did full powered sacrifice-y body throws to get me down. I did get him when he lowered his head, and I put a headlock on him and more or less sat down and scooted in at his feet, forcing a fall from him.

These past two classes were fairly neat: I feel I’m improving, and I’ve used some of our floaty techniques on people who don’t know what we do and who aren’t cooperating by providing momentum or what-not. One thing I’ve realized with the ne-waza, at least against imperfect yellow belts, is that I can try forcing his arms to cross, i.e., bring his elbow across his center to the other side, to get out of attempts to pin me. This worked fairly well. One of the black belts also showed me how to hold down one forearm if uke is on top, and then bridge on the other side to roll and get on top.

I definitely have to find a judo place in Cleveland.

Company Mail Server Upgrade

August 14th, 2004 | 11:58

Last night was the company mail server upgrade. This was basically to get Cyrus IMAPd 2.2 running on a box before I stop being a full-time employee. We used Fedora Core 2 for the underlying distribution. cyrus-imapd-2.2.3 is included as an FC2 package, which made everything easier as we no longer have to hand-roll RPMs for the IMAP server. The new box (another Dell Dimension 4600-series machine) is slightly faster at 2.8Ghz compared to 2.4Ghz, and we’re using 120GB drives instead of the 80s in the old box.

FC2 ships with openldap 2.1.29. This was actually the biggest headache, or, rather, our somewhat ad hoc schema for tracking company-specific information didn’t conform to the now-strictly-enforced object and attribute hierarchy for 2.1.29, compared to 2.0, resulting in some untidiness in the migration. Schema objects had to be recomposed as AUXILIARY instead of STRUCTURAL, and the nightly ldap dump that was used for import had to be massaged to conform. This basically took a morning to map out, but went pretty smoothly on the night of the migration following the notes that were made.

Another nuissance: the FC2 cyrus-imapd by default apparently puts its syslog messages into the “mail” signal instead of something reasonable like local6. This was fixed by getting the cyrus-imapd SRPM and editing the SYSLOGFACILITY parameter in the .spec file, and then rolling out our own slightly modified RPM. We couldn’t quite avoid doing it, but having an office package made things much easier, as this was only a two line change (the log facility parameter itself, and the package version).

FC2 also ships with an up-to-date Squirrelmail. Nice. We’ll also be better conformed to RPM updates for Squirrelmail in the future. We also have an up-to-date ClamAV (milter), though we had to go find our own milter-spamc. Snert’s version reads /etc/mail/access and uses it to exclude ALLOW domains from spam scanning, which is a good feature.

FC2 doesn’t ship with mod_auth_pam for Apache authentication, but I got a package from duke.edu after a quick Google search. It worked fine, with the following minor issues:

  • It uses its own conf.d/auth_pam.conf to load itself, so the old load lines in conf/httpd.conf had to be cleaned up to make apachectl configtest happy
  • The included /etc/pam.d/httpd didn’t work. The old /etc/pam.d/httpd did. Here are the contents of the old file:

    #%PAM-1.0
    auth required /lib/security/pam_stack.so service=system-auth
    account required /lib/security/pam_stack.so service=system-auth

After the pam.d change, it worked fine.

Also, /etc/httpd/conf.d/ssl.conf had to be edited to point at the right SSL key and cert files, even though there were directives in various virtual hosts pointing at the right place. This resulted in a confusing FireFoxerror message, basically saying that the cert couldn’t be trusted, and then preventing you from going any further. At the time, I was worried about having to go back to Thawte to get a revised cert, but I pointed IE at the site to see what would happen, and IE gave a more useful error message. This message was enough to realize that Apache was using the default dummy certs that come with the package instead of the real ones.

Those are the main points. The bulk of the time was spent waiting for IO to finish, as the mail spool was copied from the old drive to the new ones. The old drive will become a hot spare for the RAID-1 when we’re sure we don’t need anything else from it. Sadly, I didn’t have Doom 3 on any machine in the office to pass the time. During my first systems administration all-nighters at Earthweb, we had Quake LAN parties on our workstations as we waited for dark hours to start in Denver for our upgrades. And on one surreal night, I came out of our offices at 34th and Park and saw elephants marching down the street at midnight. Good times.

Arguably, this will be my last late night sysadmin campaign for a long time to come. When I got out of the office (I was considering crashing there overnight, since the couch didn’t look that bad compared to the air mattress), I was slightly giddy, and I wished I had my camera with me to take late night photos of Lower Manhattan (which still had more pedestrians and cars on it than I saw in parts of Cleveland during the day). If I did have my camera (and if wasn’t cloudy), I would have stayed up, crossed the Brooklyn Bridge, and taken pictures of the city as it caught the dawn light.

Update: There seems to be something wrong with the way CRAM-MD5 works with saslauthd. Various clients that use CRAM-MD5 for the auth mechanism fail to authenticate. This includes pine (where I first noticed a few weeks ago, but attributed it to pine itself rather than the auth library), SnapperMail SMTP, emacs sieve and possibly Thunderbird (though TB seems to fall back to PLAIN on encrypted channels). It’s a definitely a CRAM-MD5 problem because of the various clients, which are auth’ing against IMAP or SMTP, depending on the case. This isn’t a showstopper, because the main clients still work, but it is annoying. There doesn’t appear to be an updated cyrus-sasl-md5 package, but we haven’t dug into Fedora bugzilla for any reports as of yet.

Castles of Steel

August 12th, 2004 | 16:26

A couple of weeks ago, I finished Castles of Steel, Robert Massie’s history of the Great War at sea, and a follow-up to his earlier
Dreadnought, which chronicled the naval arms race between Britain and Germany up to the eve of the First World War. I’ve had bits of draft for this review lying around since then: it’s a fascinating book, colored by the personalities of the admirals and politicians (Churchill in particular), suspenseful in the description of the war’s sea battles, undergrid with the implications of command of the sea and Mahanian theories. I’ve sat on writing this review for a while, because the expanse of this book exceeds my ability to condense it in a reasonable way for a blog posting. It’s a terrific book, describing the last great acts of a form of warfare — big-gunned ships duelling in line of battle — soon to be made obsolete by emerging technologies, and the thinking, the mistakes and the luck surrounding these events.

The stage was set by the British-German naval race in the years leading up to the Great War. That race led to the two most powerful navies in the world confronting each over across the North Sea, concentrated in the British Grand Fleet and the German High Seas Fleet. At the core of each fleet were dozens of dreadnought battleships and battlecruisers, the result of a technical and conceptual revolution from a mere nine years before: the dreadnoughts made all previous battleships obsolete, and were theoretically capable of sweeping the sea of lesser vessels. The fleet with the preponderance of dreadnoughts could dominate and sink the other fleet, and thereafter command the seas. Britain had such a preponderance at the beginning of the war.

How everything played out on this stage with all the players set, is a years-long combination of accident, personality and technology. The British public expected another Trafalger at the start of the war, a decisive sea battle that would sweep the German fleet from the oceans. This didn’t happen: new torpedo and mine technology made it too costly to sail dreadnoughts directly into Heligoland for the climatic battle, and the Germans were reluctant to come out to be sunk. And so the two fleets sat on opposite sides of the North Sea, training and waiting for Der Tag. In this stalemate, the British retained command of the sea everywhere but in the sea near Germany.

Command of the sea allowed the Allies greater flexibility. Granted, some of this flexibility lead to foolishness and lethal waste in the Dardanelles campaign (though the initial naval campaign had the possibility of breaking through into the Sea of Marmara and a bombardment of Constantinople, which would have forced Turkey out of the war and restored easy communications with Russia), but also allowed Britain to detach dreadnought battlecruisers to hunt down and sink a German squadron in the South Atlantic that had been stationed in the Far East when the war began (albeit after that squadron destroyed an outclassed British squadron at Coronel, a squadron which was fatally weakened by administrative snafus in the Admirality).

Perhaps most importantly, Britain’s preponderance of naval power and command of the sea allowed it to make mistakes without these mistakes being fatal. Germany’s inferior position forced it to rely on one of two strategies, either of which had to be executed perfectly for success.

The first was a surface strategy to lure an element of the Grand Fleet into an ambush and destroy it. A successful ambush would bring the fleets into rough parity, and the High Seas Fleet could then force an action. This was risky — even riskier than the Germans knew, because the British had cracked the naval codes — and the final attempt at this led to Jutland, where the German ambushers were in turn ambushed by the entire Grand Fleet. The High Seas Fleet barely escaped, due more to poor British communications and ammunition, weather and luck than anything else. While considered a German victory at the time, because the British lost more ships than the Germans, at the end of the day the Royal Navy still retained command of the sea.

The second was the U-boat strategy, where the Germans would wage unrestricted submarine warfare on the North Atlantic. The risk of this strategy was that the United States would enter the war before the British could be blockaded into submission (while the Americans were annoyed at the British for their blockade of Germany, the British strategy — again afforded by command of the sea — resulted in lost American money rather than lost American lives). And the Germans lost this gamble, the war and their empire.

Massie does a wonderful job with the personalities of the British commanders and politicians. As an American, I’m most familiar with Churchill as Prime Minister during World War 2, not as First Lord of the Admirality during the Great War. He was a double-edged sword then, brilliant but prone to mistakes. But, as Kitchner noted, he had the Fleet ready when the war started. There’s Beatty, commander of the Battle Cruiser Squadron under Jellicoe and eventual commander of the Grand Fleet by war’s end. He was seen as the most Nelsonic of the British admirals of the time, very dashing. He also had a dark, self-serving side that wasn’t noted at the time, but once in command of the Fleet, he kept the priority of maintaining command of the sea over any sort of battle of eliminiation, no matter how desired by the British public.

And there’s Jellicoe himself, commander of the Grand Fleet from the start of the war and at Jutland, and who understood that the main strategic goal of the Royal Navy was to preserve the Fleet and retain command of the sea. Destruction of the High Seas Fleet wasn’t necessary. Actions for which he was criticized, in particular turning away from the fleeing High Seas Fleet at Jutland when threatened by torpedo attack, followed this line of thinking: as long as the Grand Fleet existed, the High Seas fleet could be bottled up and rendered ineffective. And at Jutland, he crossed Sheer’s T twice with the correct deployment of his battle line. Later as First Sea Lord, he had to confront unrestricted submarine warfare with limited resources, and eventually adopted the right mix of strategies and resources to let Britain survive and outlast the German effort.

Stereomicroscopes

August 11th, 2004 | 23:27

This past Tuesday’s New York Times has an essay on the educational benefits of stereomicroscopes. These instruments are typically used for dissection, and don’t require advanced preparation of slides. You can stick a bug in a petri dish into the viewing area, and see it magnified: for kids, this should open up new vistas, where the invisible becomes visible, yet it is not alienated from what can be seen with the naked eye. The problem with the standard toy compound microscope is that it magnifies too much, so any relationship between what’s in the viewfinder and what it was in the macro world becomes too abstract for anyone without scientific training. The stereomicroscope doesn’t have this problem because of the lower magnifying power. Plus, you don’t have to spend time preparing slides, where the process of obtaining thin slices of the subject and then coloring it with dye may be too much for a casual afternoon.

A long time ago, my parents got us one of those compound microscopes. It didn’t get that much use, I think, basically for the reasons outlined in the article: you can only do so much with proper slide preparation. And then you run out of dye, and can’t do any more. The little telescope we had got a little more use, and we did see the rings of Saturn one night.

The Times article also brings to mind all the photos of bugs that people post in the DPReview D70 forums that they took with their new macro lenses. There’s a lot of these photos, from different people: everyone has the gee-whiz feeling when taking these pictures for the first time, and sharing it with the Web. The impulse to make the tiny into something reasonably big, whether its through toy stereomicroscopes or expensive camera glass, is the same. I may have to pick up a macro lens at some point or at least a diopter filter at some point and give it a try.

Tracking Bicyclists

August 4th, 2004 | 23:10

thumbsI took a walk through Central Park in the evening and watched the bicyclists zoom through the stream of pedestrians converging on the Delacorte Theater.

These shots are with the D70 set to Auto-Focus Continuous. The focus tracking was actually pretty impressive, keeping the subject in focus from at least sixty feet out to maybe ten feet away, all along the curve of the road, while they closed in at 20 or 30 miles an hour. The lens was open to f/8, and the overcast evening light required a shutter speed of 1/10, perfect for the motion blur. This was all an experiment inspired by this DPReview.com posting on the bike race in Lower Manhattan this past Saturday.

I’m now impressed: the continuous focusing was something impossible on my old FM10 — at best, I could pick a pre-focused spot and shoot, but then I’d only have motion blur on the cyclist rather than the background. Even with the D70, the vast majority of shots didn’t come out. With this sort of tracking, I have to remeber to follow through, rather than stop pivoting when I press the shutter release. Using a rear-curtain flash might have also helped, though I’m not sure I would have wanted to distract the riders by shining a bright light in their eyes.

It was interesting standing there for the ten or so minutes I spent shooting. The cyclists do a lot of yelling at the pedestrians, who have a habit of wandering onto the road while talking on their cell phones or chatting in the middle lane, oblivious to everything going on around them. There was one near accident, where the cyclist had to swerve drastically not to run someone down. In any case, everyone should wear helmets.

Update: I just realized this cool thing. In some of the pictures, I have park lamps in the background. I was wondering why these lamps weren’t smeared across the picture smoothly. A quirk of the CCD, recording the image in pulses? No, it’s the lamps that are pulsing. You can even figure out the hertz by multiplying the flickers with the shutter speed (around 120 hertz for most of them). Neat.

Flash Sync Effects

July 27th, 2004 | 13:39

There’s an interesting discussion in this DPReview thread, where an apparent Photoshop hack job — a cut-and-paste of a skater jumping over a fence — turned out to be a real shot. The give away weird shadowing — why would there be a shadow against the background sky? — turns out to be what you get when a front- or rear-sync’d flash is fired to stop a fast moving, silhouetted object against a relatively bright background.

Basically, the flash is illuminating the object for only a portion of the time the shutter is open. For the rest of the time the shutter is open, the CCD/film is capturing a dark object moving against the bright background. The shadow in the sky is actually the motion blur of the silhouette.

The shadow will appear before or after the object, depending on the direction of motion and whether the flash is fired at the front or rear of when the shutter is open. Obviously, in front of the object with the front sync, and behind with rear sync.

This is just a lesson in counter-intuitive results from the camera.

Note that rear sync flash is recommended for fill flashing rear illuminated subjects, e.g., a person in the foreground, and a brightly lit Empire State Building in the background. The shutter has to be open a relatively long time to properly capture the background (the exposure is metered against it), and the flash should go off at the end of this period, because people take the flash as a cue that the picture has been taken, and may start to move. A front sync may therefore result in a blurry subject because he started to move before everything was done.

New Camera

July 22nd, 2004 | 11:27

I got the Nikon D70 last week, and have had time to play around with the controls, as well as getting the necessary protective kit, like a UV filter for the lens and a basic bag for the camera itself. I spent last week getting used to the thing: the controls are just very different from the fully manual camera I had before. How does one set aperture and shutter speed? How does one use the exposure meter? What’s this “white balance” thing? How does one vary a bit from the default exposure? Lots of shots down my empty apartment while fiddling with the buttons and dials (one hallmark of DSLRs as opposed to digital P&Ss is the ease at which you can change settings, rather than drilling down half-a-dozen menus to change the aperture setting). But no matter: the marginal cost for each of these pictures is practically zero, which can’t be said for film.

Here’s a few shots from my first “real” set, walking up West Side Highway and Hudson River for a little bit:

first photos

The bunch in the middle (DSC_0017 through DSC_0023) was one of those experiments that require a bit too much discipline for me with my film camera: I was adjusting nudging the EV from a stop below to a stop above the camera-determined optimal exposure in 1/3-stop increments. With the film camera, there’s no instant feedback, and there’d be a cost in doing this (albeit a couple of bucks for film and processing).

There’s some aliasing in the sky with a number of these shots, but I think this is an artifact of (Canon-oriented) BreezeBrowser’s NEF to JPEG conversion than anything else: I don’t see the aliasing when looking at the NEFs in BreezeBrower’s “high quality” view. I probably will wind up getting the Nikon Capture 4 software sooner rather than later, since NEF processing is that NC4 is meant to do. NC4 will also alow control of the camera through the USB cable, as well as loading in custom curves that may give slightly better in-camera JPEG output than Nikon’s defaults.

The camera itself isn’t particularly heavy. It’s a bit bulkier than my manual, but that might be cause I’m using a heavier lens right now, before I get the 50mm f/1.8D. I feel like I’m actually photographing things with this camera, unlike the P&S Sony I had been using (I never got perfectly comfortable with looking a the LCD at arms distance to see the shot).

In the meantime, I’m learning what I can. The dpreview.com D70 forum has been very helpful, though the photos people post tend to get too many oohs-and-ahs as opposed to proper criticism, and a bit too much focus on weird minutiae. Besides the forum, there’s Thom Hogan’s guide to the D70 and this apparently more general book on digital photography. There’s also the Nikonians forum.

In terms of settings that I can set without the aid of NC4, there are these that can be changed or at least considered. Some things I should do also is set the in-camera to “none” and do any sharpening in post-processing.

Apparently, there’s much to figure out. It’s a nice camera. I’ll be using it in August, being a tourist in my home town before heading out to Cleveland for good.

Oishi Judo

July 19th, 2004 | 13:06

My teachers at Eizan Ryu jujitsu encourage their brown belts to go out into the world and look at different dojos, different arts, under the assumption that no one art or style has complete answers and it’s best to see how other people do it. Cross-training is good.

This past Saturday, I took my first judo class at Oishi Judo down on Leonard Street. I figure that a few judo would be best right now, since I’m going to wind up primarily in aikido when I permanently move to Cleveland, and Oishi isn’t too far from work: I can maybe take a long lunch on Wednesdays and take classes there.
Read the rest of this entry »

Fourth of July Weekend on Lake Erie

July 18th, 2004 | 22:00

I flew into Cleveland on June 30th, taking an early train from Penn Station to Newark to catch a Northwest flight to Detroit and then to Cleveland. The AirTrain got to the airport early, which was a good thing, because my credit card didn’t pull up the flight on the NWA kiosk for the e-ticket. The customer service rep at the counter told me that I wasn’t on any NWA flight, which was surprising. Digging into his systems a bit deeper, he discovered I was actually on a Continental flight through code-sharing: I’d have to go to a different Newark terminal to check in. But the rep was helpful. Seeing how early it was, he put me on a NWA flight that would leave half an hour before the Continental one, and he could give me the boarding pass there.
Read the rest of this entry »