John Patrick writes: The Spam Has Got To Go. Amen to that.

But what's the solution? There are a lot of things you can try: legislation, informal sanctions, blacklists, spam filters, hashcash schemes, p2p spam reporting networks, etc. Many of these are being tried. Jon Udell and others suggest that digitally signing emails will cut down on spam, but it seems to me just one more barrier for spammers to overcome. Tap into an unlimited supply of digital id's, and you just send signed spam.

I think I know how to stop spam, while preserving the wonderful openness of email. Longtime readers won't be surprised to hear that my idea is based on attack-resistant trust metrics. In fact, longtime readers are probably sick and tired of me harping on this damn trust stuff all the time. But harp I will, for the message needs to get out there. There is now research on how to build spam-resistant messaging networks. Quite a bit more work is needed to make it practical, but I think the usual technique of paying smart people to think about it has a good chance of working.

Anyone who's interested should read Ch. 7 of my thesis-in-progress.

SourceForge

I had a nice interchange today with Jacob Moorman of SourceForge, about my miffed feelings regarding the language on their peer certification page. Actually, the peer certifications are going away, but more important, I get a renewed sense of an open channel with the SourceForge guys. In fact, complaining in my diary rather than contacting them directly was unfair, and I apologize.

I really hope SourceForge manages to succeed in spite of the current harsh business climate. They're doing an immense public service to the cause of free software.

A New Kind of Science

I power-skimmed the book again tonight. There's no question about Wolfram's ego. There's precious little actual science in the book, but I'm still convinced that there's something there. Exactly what, I'm not quite sure. My latest stab at expressing his central point is: "a lot of shit is universal".

Even so, I'm not really swayed by his arguments. One of his points seems to be that people thought about universal computation ability as the exclusive province of very complex, specialized equipment, while in fact, many simple systems can express universal computation. Sure, sure. But his Turing machine construction on rule 110 requires enormously complex and specialized software (the initial state), even if the hardware is ridiculously simple compared with previous universal constructions. If you look at the sum complexity of hardware and software, then there's a good argument to be made that Wolfram has taken a big step backwards.

Distributed proof repositories

AaronSw writes: "While hashes are useful, one thing you often here in W3C circles is that local file storage should keep track of the URIs from which it got the files and cache them so. So you'd just look for the cached file with that URL instead of the one with that hash."

Glad to see the idea is resonating for other people. I'm not sure what UR? scheme is best for these proof files, but my hunch is that hashes will be important. For one, definitions, theorems, and proofs are probably a lot more immutable than most of the other objects W3C people deal with. I'm also envisioning a Napster-style propagation of referenced objects. In typical usage, I can imagine a person working on a proof in conjunction with a client which snarfs proofs and lemmas from the rest of the Web, then, when it's done, uploads everything to a static web space. Thus, URL's move around fluidly, but the hash is constant.

I think I'll write about choice of primitives over the coming days, unless people beg me to stop.