David Gerard

The marketing sense of the word “evangelism” was popularised by Guy Kawasaki in his time at Apple. Guy wrote a book on the subject, Selling the Dream (Amazon link) about how he did it at Apple, which I highly recommend. Guy also blogs chronically about this sort of thing. I need to find my copy again … its in a box somewhere. (My house: “Where’s x?” “It’s IN A BOX!” One day everything will be unpacked … then we’ll probably move.)

Read this post: The Art of Evangelism. Apply all ten steps to Wikipedia and Wikimedia.

(This suggests to me that free software “gateway drugs” really work, e.g. Firefox, OpenOffice, GIMP. Users care about applications; once they’re using all-free applications, swapping the OS out from under is easy and they sudenly discover their battery life has doubled from not running an antivirus. Sysadmins are already used to swapping Windows out from under a stack, putting Linux in its place and vastly improving performance. We use Wine for this on business-critical systems at work. This suggests that Microsoft’s drive to make Windows a first-class platform for open source software will in fact shoot them in the foot. I’m sure they have a game plan that says it won’t, but I still can’t see what it might be myself.)

You know you’ve made it when you get onto the special mailing list that includes the FBI, CIA and White House. (The place, not the band.)

DANGER!
Clicking here may induce
bleeding from the eyes.

“Can you imagine what work life would be like if one of the conditions for promotion was you had to give away everything you knew to people who could use it for their growth and development, and you had to reach out and help people to be successful, and you had to demonstrate you were open to being helped by others in your own pursuits.”

FayssalF suggests that the sentence “We agree to make this a place where we extend a hand to each other” be put at the top of every talk page.

Knol is Google trying to recreate Squidoo or Helium, not an encyclopedia. Wikipedia is #8 on Alexa, Squidoo is #431, Helium is #4999 and only Google knows how well Knol is actually doing. I mean, I was incredibly impressed when I first joined Wikipedia in early 2004 that it was #500. But nevertheless. At least about.com makes #86.

(In fairness, Google has never pushed Knol as a Wikipedia killer; that’s entirely a media-created synthetic controversy.)

There’s hardly a “Wikipedia replacement” that hasn’t started from trying to make a welcoming environment for authors. Wikipedia, however, is popular because it’s what readers want. Writers are important, but way less so than the readers.

I’ve seen very few Wikipedia replacements or even forks that aim primarily at creating a better resource for the reader, and leave the rest to happen. Citizendium is the only one that springs to mind — CZ is very reader-oriented, and slowly accumulating lots of good stuff. It also expressly tries for good writing, unlike Wikipedia.

If readers wanted ten articles on one topic, they’d just click the first ten Google hits. It’s like metasearch engines that gave you results from ten bad pre-Google search engines in the hope you might find a damn thing, when the real answer was one search engine that didn’t suck. Tell you what, the main value of Cuil is to explain to the kids how bad search engines were before Google got it right. One good resource kills ten mediocre resources.

Leaving the editors to battle it out to collaboratively create the one article on a topic appears to have worked to give readers the simple quick reference site they actually want to use. Inherent unreliability and all. Discuss.

I’m sitting here with my dear friend Kirrily Robert of Freebase. Her office is being remodelled, so decided to work from home in London for a week. We hung out with the geeks and drank to excess on Sunday (Kirrily says she drank to “sufficient”), so today we’ve been geeking Freebase and Wikipedia and social content creation and so forth.

Freebase is a collection of structured data, with little or no notability barrier. (Spam is fine if it’s structured data!) The differences from Wikimedia are that (a) it’s all CC-by (b) it’s run by a company, not by a charity. The differences from Google Base is that (a) you can do mashups of every data table with every other data table (b) they don’t want your private data (unless you want your daily calorie counts available forever under CC-by).

I didn’t think it was way cool until she showed me David Huynh’s Freebase Parallax demo video. I most strongly urge you to watch this.

Advancing Freebase is in line with Wikimedia goals, as it’s useful free content (and the dumps work). The really good thing you can do is: if you’re getting someone to release a bunch of data, do your damnedest to get it under CC-by or public domain. That way we can have it and they can have it and everyone can have it.

The other thing we rambled about was the social structure of the thing. At the moment Freebase’s Alexa rank is about 47,000; socially it sounds like Wikipedia in 2002. The key point is that in a public participatory content production project, people are all your problems and this is not susceptible to quick fixes, technical or social. Just so she knows what they’re in for.

London readers: there’s a Freebase meetup at the Yorkshire Grey pub in Holborn from 6:30pm.

Fuel your editing! Keep a tall, cool glass of Wikipedia juice to hand. Just the thing with stir-fried Wikipedia, congo eel with Wikipedia or just a couple of slices of Wekipedia toast.

Ogg Theora and Ogg Vorbis support for the HTML5 <video> element has landed in Firefox Minefield nightlies (3.1a2-pre). This is big news because it means a standard way of displaying video in web browsers will be available to all without being stuck with Flash. And Theora is the only accepted format on Wikimedia Commons. Posts: Greg Maxwell, Christopher Blizzard, Chris Double, Gervase Markham.

What we need is people to test this. So please download a copy of Minefield, test it thoroughly on Wikimedia Commons video, beat on it, thrash it, report bugs. There’s plenty. You need to load the video, click “More …” and it’ll give you the option. Wikimedia would very much like to make it a first option rather than a last one, but first it needs to be better (more functional and stable) than loading Cortado with Java.

Apple and Nokia tried some truly disgusting FUD around the topic and successfully got the words “Vorbis” and “Theora” taken out of the HTML5 spec, but Firefox adoption means 20% of Web users in short order. So we can leave them to play catchup per business needs. “You got a Nokia? No wonder you can’t watch that Wikipedia video, Nokias suck.”

The Wikimedia blog is still on WordPress 2.5. I moved three blogs on my own site — this one, Cyber Chatelaine and Rocknerd — to 2.6 today. The first two were fine because they were all but unmodified from the standard install; the first has lotsa tweaks and extensions, and I had a marvellously annoying time this afternoon fixing it up and I’m still not finished. Grah. I have recommended severe beta testing for the Wikimedia blog before updating.

Note the early start time! 11am. (So people going to Wikimania can catch planes.) Signup page. Arkady Rose and I (and the small child) plan to attend, particularly with the intent of seeing how to make Wikimedia UK more useful for something. I really should attend more of these …

FritzpollBot was recently approved to create stub articles on English Wikipedia for most or all of the documented villages and towns in the world. (Example.)

In 2003, Rambot created placename articles for every census location in the United States. We were therefore able to claim complete coverage (per that “encyclo-” prefix) of one country. FritzpollBot aims to complete this coverage for the entire world.

I think this bot-assisted programme of article creation is a Good Thing for topics where we do in fact have the data. It’ll certainly help alleviate our systemic bias. The issues I can see are editorial — the Rambot articles are data in prose form that these days we’d do with a parameterised template, etc. — but Fritzpoll is quite aware of these and the planned programme includes considerable human review and the active involvement of country WikiProjects. Good.

(May I note that people whose objections are that this will artificially inflate the article count or make Special:Random annoying appear to have forgotten that we’re here to write an encyclopedia.)

The question that springs to mind is: what else can we get complete data on for bot-assisted article creation? Every state-level or higher politician in every country ever? What else?

Update: Fritzpoll is proceeding with all due caution, and the bot will be doing nothing but preparing lists as yet. See evolving FAQ.

If you ever wonder why you bother working on Wikimedia projects: 1, 2.

Has anyone ever gotten a straight answer from licensing@fsf.org about GFDL queries? I have never even heard of an answer from them that isn’t their Magic 8-Ball imitation. “Reply hazy, read the license text and ask your own lawyer.” Our lawyer is Mike Godwin and he says it makes his head hurt. YOU WROTE THE DAMN THING. WHAT DID YOU MEAN? WHAT WERE YOU THINKING? ANSWER ME!

In fairness, the FSF contact page says licensing@fsf.org will help with “questions about the GPL and free software licensing.” Even the FSF has given up trying to make sense of the GFDL. The new version can’t happen soon enough.

(Provoked by asking for help with the reuse FAQ and the likely utter unfeasibility of audio versions of GFDL text. The latter is one of the best arguments I can think of for running screaming to CC-by-sa as absolutely soon as possible and throwing the GFDL into a fire.)

Last Thursday at London.PM, I got asked a lot why MediaWiki wikitext doesn’t have a WYSIWYG editor. The answer is that a WYSIWYG editor would need to know wikitext grammar, and there is no defined grammar. The MediaWiki “parser” is not actually a parser — it’s a twisty series of regular expressions (PHP’s version of PCREs).

So any grammar effort (and several What You See Is All You Get editors — others just forget wikitext and write HTML) requires reverse-engineering that, and lots of people have tried and gotten 90% of the way before stalling. It doesn’t help that wikitext is (I’m told) provably impossible to just put into a single lump of EBNF.

The goal is to replace the twisty series of regexps with something generated from a grammar. Tim Starling has said, more or less: “We can’t change wikitext. Go away and write something that (a) covers almost all of it (b) is comparably fast in PHP.” Harsh, but fair.

It occurred to me that there must exist tools to convert regexps into EBNF. And that if we can get it into even a few disparate lumps of hideous EBNF, there should be tools to take those and simplify them somewhat. (Presumably with steps to say what given bits mean.) Or possibly things other than EBNF, just as long as the result is parseable.

I am not (even slightly) a computer scientist, but many of you are. Does anyone have any ideas on this? Or pointers to anyone having done anything even remotely similar? Or knowledgeable friends they could point this query at?

The other approach is parserTests.php. Running maintenance scripts, the scripts (look for parserTests), the list of tests. A “parser” will be anything that passes the unit tests.

Germany’s Brockhaus Encyclopedia Goes Online.

Wikipedia gained its present hideous popularity through convenience — an encyclopedia with a ridiculously wide topic range, with content good enough to be useful no matter how often we stress it’s not “reliable” (certified checked) as such.

Britannica and Brockhaus may be theoretically higher quality, but are not right there on everyone’s desktop — they fail on practical availability. Worse is better. Most of Wikipedia’s readers (the people who make it #9 site in the world) wouldn’t have opened a paper encyclopedia since high school. Wikipedia fills a niche that was previously ignored when not botched.

So the paper encyclopedias put their content online. Can they provide a better website than Wikipedia? Ignoring the process, just looking at the resulting body of text? Can they produce content on the range of topics people look for on Wikipedia fast enough at their advertised quality level and keep it up to date? To what extent can they compete with Wikipedia without becoming Wikipedia? What would that entail?

“Really, I’m not out to destroy Microsoft. That will just be a completely unintentional side effect.”