28 Jul 2004 Bram   » (Master)

I still have my domains

Thanks to everybody who mailed me translating the mail about my domains. A whole bunch of people translated it for me, and David Turner went so far as to call ovh and tell them about the apparent hijack attempt. And Joey DeVilla got in contact with James Woods at Tucows.

All the help turned out to be way more than needed. I probably could have ignored the original mail and not gotten my domains taken, since it required affirmative action on the part of one of the contacts listed, and for those two I was the only contact.

Right before posting here about my domains I mailed tech support at my registrar, domainmonger. A few minutes later I noticed the domain locking feature, and turned it on. Somewhere about that time I got mail from tech support at domainmonger saying that they'd locked all of my domains. I'm not sure whether they or I did it first.

Thanks everybody for your help, that was confidence-inspiring.

In answer to the obvious question 'Why didn't you just hit the site and say no?' First of all, the site was in french, which I only speak a few words of, and second of all, when I went to it it said that I needed to enter some code taken from my registrar. While that message was probably legit and there for security purposes, it set of an alarm bell in my head that the attempted transfer mail I got might have been a fake sent by a spammer in an attempt to get me to give them a confirmation code to steal my domain. So I didn't do it.

Cool Mechanical Stuff

Here , here and here you can find lots of explanations of cool mechanical things.

My own crackpot engine ideas

I've designed a piston replacement I call a 'steel lung'. It basically acts like a piston in that it contains a chamber which expands and contracts, but unlike a piston the shape of the chamber remains mostly spherical for the entire stroke, for much better surface area to volume ratio than found in conventional pistons. Of course this creates extra friction, although not an insane amount - the amount of 'seam' is only two or three times what you have in a piston. It also increases sealing problems, but again not as much as you'd expect. It can be configured so that the expansive forces push the moving pieces into each other rather than away from each other, and the sealed edges move outwards during combustion, so there isn't all that much force on them.

The big question, of course, is whether heat loss during combustion is significant enough to be worth reducing. I'd appreciate it if anyone has any knowledgeable opinion about this.

Another idea of mine is based on how the heat of the piston head is frequently a problem in car engines. How about if, rather than a conventional piston, we have a piston which is composed of several sliding pieces in such a way that the piston head is actually a sliding pieces whose exposed part is different with each ignition? That would cause the effective surface area of the piston head to be a lot larger, thus making it easier to keep cool. Again, I'd be interested in any knowledgeable opinions about whether this is a worthwile goal even in principle. I have a design for this, but it lacks elegance.

Finally, I have a question about the way in which motion is translated from reciprocating to rotational. It seems a little odd that real engines do this with a simple camshaft, since that has a very particular rate of movement at each point, and I have a hard time believing that it's optimal. There are plenty of only slightly more complicated mechanisms, many given in the links above, which can modify this with a lot of precision. Does the relative rate of movement of the piston at each point in the cycle not matter all that much, or is it simply not worth the extra complexity?

Buckminster Fuller was a crackpot

Buckminster Fuller's intended great invention was the dymaxion car, which was basically a stretch limo with one front wheel which always ran backwards. Its funding was pulled after a fatal crash.

You can read Fuller's excuse-making about the crash here. For those of you unfamiliar with the physics involved, I'll explain. The car turned left and, as is inevitable with a 20 foot long vehicle with real-wheel steering, its butt stuck out into the right lane. There happened to be a car there (Fuller claims it was 'rubbernecking', which is completely besides the point - this is a very common kind of accident which shouldn't cause a fatality). When the rear side of the dymaxion hit the other car, its single rear wheel got shoved hard into the ground, making it precess and forcing the turn to be even greater, quickly causing the car to get to such an angle that it rolled over.

The investors wisely decided to pull out after this, presumably because they wanted nothing to do with a company with such callous disregard for passenger safety, and that was the end of that. Fuller did spend a huge amount of his own money attempting to build more cars, but they never went anywhere, probably because the quality of his engineering ideas was about what you'd expect, which is to say, very poor.

When you look at Fuller's accomplishments, philosophy, and claims, they bear a notable resemblance to many internet pundits today.

Moller Skycar

Speaking of which, those of you who have heard of the Moller Skycar may wish to read this.

If you're interested in flying recreationally, I recommend powered paragliding.

Simplified Poker

When musing about poker, I came up with simplified version, which is close to as simple as you can get. Two players, three cards. Each player is dealt a single card. The higher card wins. After anteing up, each player can either fold, call, raise one, or go all-in. I suspect that this variant would actually be fairly interesting to play, since it involves going all-in (which I think makes complete mathematical analysis completely beyond the capabilities of current techniques) and the small number of cards make the contents of the other player's hand strongly related to the contents of your own.

An improved trust metric

I think I've figured out how to improve on the last batch of trust metric ideas I posted. In those, I had problems with everything certified by the 'root' looking like spam, and suggested some very kludgy ways of dealing with that problem.

A much better approach is to view spam guilt as 'leaking' out of the system, an extremely small amount from everywhere. I'll present a technique as an interative process, in which spam guilt gets back propogated with each round. After many rounds have been done, it will be clear that some nodes have their guilt score increasing without bound, while the rest will have their guilt scores stabilized. All nodes whose guilt surpasses some threshold are then removed (the number of rounds and threshold are obviously a judgement call, but generally more is better). The distinctions between directly vs. indirectly attacking nodes I posted about before are then made, and the decision of which nodes can mail is based on the result.

(At some point I'm going to post a consolidated form of my best trust metric to date, so people don't have to read a long series of posts and cull the non-obsolete snippets out of them to figure out what's going on. In the mean time, I apologize for the vague references to stuff I've posted about before, and please bear with me.)

First, we have some value for how much spam each node is allowed to send before getting cut off, which we set by judgment call. A hundred pieces sounds reasonable, so I'll use that for the sake of argument. Next, we calculate how much 'leak' each node is, which we can reasonably set to 3 * (total amount of spam ever sent) / (total number of nodes). The number 3 can obviously be varied, but that's a good value.

On each round, we add spam score to direct infringers, back-propogate spam scores and then leak spam scores. We add spam scores for each direct infringer simply by adding the amount they've spammed to their score. To back-propogate, we iterate over each node, and for each of them we find the minimum spam score of it and any of the nodes which certified it, and transfer 100 to that node. If the node in question's spam score is less than 100, we transfer that amount. To leak, we reduce the spam score of each node by the leakage amount, and round up to zero.

I'm quite happy with this algorithm. It's essentially linear time, and I can straightfaced say that it's both secure and practical.


bgeiger: If one were to select a game to be the single canonical abstract strategy game which people the world over knew how to play, having draws would be a complete disqualification. Go has no draws if you use a super-ko rule and non-integer komi, but there are many modern board games which meet that criterion as well, for example Amazons.


Enough blogging for one sitting. More later.

Latest blog entries     Older blog entries

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!