Slashback: Pop-Ups, Books, Qmail

Slashback tonight is loaded with updates and addenda to previous stories on Bayesian spam-prevention, pop-up ad blocking, and celebratory picnics as well as an inquiry into the other side of visionary literature. Read on below for the details.

What's your idea of feel-good literature? A few weeks ago, an Ask Slashdot question was posed about the greatest dystopic novels, and quite a few people weighed in with their choices for visions of the post-nuclear, post-germ-warfare, post-natural disaster or otherwise blighted future.

Now reader itwerx wants the other side: "That "Dystopic novels?" Ask Slashdot was so darn depressing we need a counter balance! Let's hear what novels of utopia may not be widely known."

It's certainly widely known, but I'll start the bidding with Atlas Shrugged.

The best revenge is living well, and gluing spammers end-to-end. RealDhar writes "Hey, just thought I'd let folks know that, inspired by the recent article about Paul Graham's Bayesian spam filter work, I went and wrote one for qmail. Please check it out!"

What took so long? Pop-up ads are no fun. iVillage cut them out, AOL swears they're cutting back, and even Netscape 7 can be wrangled to block them. An anonymous reader writes "From the Associated Press (via Salon): EarthLink Inc. said Monday it plans to offer its subscribers software to block Internet pop-up advertisements as part of a wider campaign to set itself apart from competitors. The full story is here.."

Penguins and picnics go well together. ArtEnvironment writes "Besides today's 2nd California Linux Anniversary Picnic previously mentioned, there will also be PLUS, the Philadelphia Linux/Unix Symposium which is the 2nd annual East-Coast Linux anniversary picnic and more, including a bar night kicking off Friday the 23rd, a free computer/electronics swap meet and giveaway on Saturday the 24th, and of course the picnic on Sunday the 25th. Also included is one of the well-known PLUG GPG Keysigning parties. PLUS will be an annual grass-roots event, but it 'won't be big and professional like' ALS or LWCE. ;)"

I look forward to the final, triumphant mention of this :) Qbertino writes "The Blender Fund, established a month ago in order to buy the IP of the 3D Pakage Blender and, at last, GPL it, has accumulated 90K Euro (90K$) of the required 100K in less than 4 weeks. As it indicates on the Website, Ton Roosendahl, father of Blender, is preparing to release the sources which should happen within the next week or so. Time for a Blender icon on /."

Comments

Blender and Free fonts

[PeterClark]

To merge the topic of Blender with that of another recent subject, has anyone started a fund for creating Free fonts to eliminate Free software's dependence upon Microsoft's fonts? From the discussion that has already occured, it seems as though the only sane and reasonable way to get high-quality, consistant fonts is to scrap some money together and pay a professional to do so.

People, if a _rendering_ program, that is probably used by a relatively small amount of people, can reach 90% of its goal in four weeks, what can we do about raising funds for fonts, which everyone has an interest in? What we need now is for someone or some organization well-respected within the community to speak up and say, "The pot is open! Come chip in!"
:Peter

On a related note...

[quinto2000]

This idea of using Bayesian filters on Spam reminds me of an idea that I've tossed back and forth a couple of times with other slashdotters. I'm getting into neural nets, and I think it would be really interesting to do this kind of analysis on one of the largest data set around, the Slashdot comments. It would be a perfect database for training, because user moderations are attached to most comments. Specifically, I think it would be really cool to train a neural network to recognize trolls. Has anyone else ever thought about this? I would even be able to get academic credit for the research, but CmdrTaco didn't like the idea when I suggested it to him.

Just like the argument for bayesian analysis of SPAM, reason-based analyis of trolls is fundamentally flawed, as can be seen by the broken "lameness" filters. A neural network/bayesian approach would probably work much better at finding the features trolls have in common. Slashdot could mark likely trolls automatically after they are analyzed by the system, and users could filter "likely troll" in their user preferences page. But mostly, this would be a cool project to do, and I wish CmdrTaco would be more willing to allow direct database access for academic projects. Screen-scraping is not an attractive prospect.

Re:On a related note...

[yog]

Well, you're proposing to do something really hard, i.e. to somehow
recognize trolls among legitimate messages, which sounds nearly
impossible, and yet you balk at doing something extremely easy,
i.e. cutting and pasting messages into an editor for a few days or
weeks. You'd be doing the world a huge service if you could solve the
troll/spam problem; go for it. Don't let lack of direct access to a
database slow you down.

Not all fonts have to include Chinese

[yerricde]

Before anyone here says that fonts are easy to make, you're probably forgetting the non-western character sets and the thousands of unicode characters.

Just as there are fonts that specialize in "CJK" (Chinese Japanese Korean) glyphs, there can also be "LGC" fonts for Latin, Greek, and Cyrillic. A good font editor will have the user draw a bunch of glyphs representing A-Z in Latin, the Greek, Cyrillic, and IPA glyphs that do not match the Latin glyphs, and then some diacritics. Then from that data, it'll "compose" glyphs for the first 1500 or so characters in Unicode.

Another optimization: when creating a new glyph, copy parts from similar glyphs and present them to the font designer for further work. For example, from b and p, you get (thorn). From D, you get Ð (edh). From l and n, you get h. From n, you can infer most of m. From f, you get long s, and from long s and normal s, you get ß.

Re:Not all fonts have to include Chinese

[foobar104]

Um. That actually sounds like a terrible way to design a font. I've never done any font design myself, but I used to work for an agency that employed a guy who could. He was no Twohy or Slimbach or anything, but while I was there he designed three Latin-alphabet typefaces. He drew each glyph by hand on big sheets of paper, then scanned them and converted them to vector art. He spent weeks trying out tons of combinations of letters to see how the forms would look together, designing ligatures where it was appropriate. (A ligature is a single glyph that looks like two letters. The classic example is the "ae" ligature which looks like an "a" and an "e" smooshed together, but that's not terribly common. Every professional typeface, though, has ligatures for the letter combinations "fi," "ff", "fl", "ffi," and "ffl." Look at an "fi" in a book through a magnifying glass. You'll see that the bar of the f is connected to the i, and that the dot above the i is absent. It's a ligature.

Anyway, my point is that designing usable general-purpose fonts is a lot more work than you imply here.