Debian Developer Center Of Mass

Edward Betts writes: "Debian One is over, we are at LSM, and it is raining, what do we do? Try and decide the location of the next Debian conference of course, and we all know that the best place for a Debian conference is Debian's centre of mass." What an ideal location for a conference -- perhaps they can devise a mission to retrive the errant U.S. hydrogen bomb (more information too).

Comments

Assumes developers all weigh the same.

[hatless]

Since Americans weigh more than everyone else except maybe some Pacific Islanders (of which very few are Debian developers), this study should have taken that into account, especially in light of the significant number of Scandinavians and Finns in the European contingent.

I suspect that if this were taken into account, the conference would best be held a couple hundred miles northeast of Newfoundland.

Better measure by travel time

[streepje]

If you could get any kind of a fix on travel time rather than distance, this could be useful.

Cost would be even better.

Besides, the real center of mass is somewhere way underground.

not welldefined (a mathematical note)

[Anonymous Coward]

There is a problem when calculating center of mass on a sphere. It's not really welldefined. Imagine e.g. that all the developers where distributed evenly all over earth. Then everyone could claim to be at the center. Another problem is that center of mass is based on vectorspace properties such as identifying points and vectors. This is not possible on general manifolds (including spheres) The result from making averages over coordinates depends on the coordinate system and is therefore not welldefined in a physical sense.

phew!, I hope this saves the debian community the hassle to swim around in freezing cold water during conference....

Re:not welldefined (a mathematical note)

[jonathan_ingram]

The research project I am currently engaged in involves finding 'averages' on spaces that aren't even manifolds. It's very easy to do on spaces that are uniquely geodesic, and I'd imagine you could get a plausible measure on spheres as well. All you need to do is to be able to measure distances between points -- and a sphere is certainly a metric space.

Just find the point (or points) which minimize the sum of the squares of distances to other points. Because the sphere isn't uniquely geodesic, you might get more than one point (if you had two points, one at the north, one at the south pole, then the points of minimum squared distance would be all the equator).

Re:not welldefined (a mathematical note)

[PD]

Remember that the Earth isn't a perfect sphere, so the problem of having each developer being equally spaced might be such an unlikely occurrence that we don't need to write code to handle that special case. Instead we could just put a note in the manual that in the event that the Earth becomes a perfect sphere and all developers are equally spaced, the program will segfault. The workaround would be to take a step to the right, and then restart the program.

Um, No.

[John Murdoch]

Um, sorry....

There isn't going to be a Third World War. And the Internet is the reason why. When the masses can--at trivial expense--discover what they choose from wherever they choose to find it, the power of governments (including those spouting the rhetoric of Lenin and Engels) is demonstrably weakened. The Internet really does bring "power to the people."

Power to the people. Right on!

But the deficiencies of Lenin and Engels are sort of off-topic, so let's rein in our political diatribes just a bit and look at a slightly different point: you're making a whopping great logical mistake. Just because 90% of the Debian developers live north of the equator doesn't prove that the "billions upon billions of...unwashed masses" don't have access to the Internet. Note that the concentrations of Debian developers are also heavily distributed toward parts of the world where most educated people speak English (with the notable exception of India). You're also assuming that Debian developers are evenly distributed across the Internet--but there's nothing to prove that's true.

Analogy: what results would we conclude by doing this same pseudo-analysis of the FetchMail developers? I'd bet we'd discover that they're disproportionately close to Chester County, Pennsylvania, and that lots of them own guns. Why? Because ESR lives in Chester County, Pa., and is something of a gun nut. Developer communities tend to be social communities--so if you're a pistol-packin' programmer, you're likely to be among friends on the FetchMail project. That does not mean that Internet programmers in general, Mail subsystem programmers in particular, or even American programmers are disproportionately armed. It's a self-selected cluster, and you can't draw valid conclusions from the traits of self-selected clusters.

Geography doesn't matter, airfare does...

[John Murdoch]

Hi!

I'm a GIS developer, so I'm just as happy to geocode data points and map them as anybody. Party on, you geo-coding dudes! But the "center" you have defined assumes that you're traveling "as the crow flies"--and (pardon the old joke) that's only useful if you're a crow. If you're really serious about coming up with a logically-derived meeting location, central to as many people as possible, I'd suggest a slightly different method.

• First, identify major international airports in the areas where you have Debian developers.
• Map those points, and then identify (for each Debian developer) the two or three closest airports. (For developers in the more remote regions of the world you might need to just deal with the one choice they have.)
• Then identify a dozen or so centrally-located international airports. Don't immediately pick Heathrow, DeGaulle, and JFK! In particular, consider Shannon (Ireland) and Anchorage (Alaska)--both see a lot of air traffic from very long distance flights, and have an amazing number of connections to/from practically anywhere.
• Then identify median airfares from airports to these potential central locations. I suspect that you'll find that airfares in Europe are higher than fares in the U.S., and fares south of the equator are higher than fares to the north. This should "tilt" the scales toward a European location.

I'd bet that this wouldn't take that long to figure out. You've probably only got 20-25 airports to check, and using any of the travel sites you can shop for fares in very little time. You're not going to come up with a single solution: but you're going to narrow your list down rapidly to just a few choices--you can then consider other factors (how expensive hotel, food, and rental car expenses will be; costs for conference organizers to bring keynote speakers [since the conference pays for those], etc.; whether suitable space for the conference is available on your dates) and make your choice.

And, oh, yeah--where does everybody want to junket to next year? Even if Shannon, Ireland is the ideal location, you can't have the conference there every year....