Saturday, November 11, 2006

A new podcast. I've been flashing on all kinds of things after Tuesday's election. We've never had so much power. We sure didn't get fooled again. A branch of government switched parties, and all of a sudden we got our democracy back. What's next? 

NY Times: "The president and his political adviser, Karl Rove, refused to compromise on legislation, bullied their own party's senators and ignored leaders of the opposition." 

Does hype ever go out of style? 

Just when you may have thought it was safe again to make software for users without so much hype, John Markoff, writing in the NY Times, says Web 3.0 is "in its infancy." Who says it's the next thing? Hmmm. Hard to figure out.

Stop for a moment, take a breath. Do we really want to go here? Just as Web 2.0 is petering out (we hope) can we avoid the next bit of bullshit?

The hype habit of Silicon Valley is pretty bad. How about setting expectations somewhat in line with reality? Sell what you have, not what you're dreaming about. Every time SV pumps up expectations, the money starts flowing into the hype, and away from what keeps people employed -- the technology SV is supposed to be producing.

If Hollywood worked this way, they'd sell the movies they think they're going to ship in 10 years. If Nashville worked this way, they'd sell the country music of 10 years from now. If NY worked this way, they'd only sell futures, you wouldn't be able to buy stock.

More hype should be as welcome by the technology industry as another Karl Rove slogan is welcomed by today's electorate (hint: the president's approval rating is down to 31, a drop of ten points in less than a week).

Danny Ayers: "Who else is lumping these things together and applying that label?"

Tim Finin: "The article is pretty much content free from a technology perspective."

Postscript: The features they're hinting at are hardly new. Collaborative filtering is a mature technology, although there's certainly room for innovation, and a lot to be gained by opening up the silos. Example: Netflix movie recommendations. They know which movies I'm likely to like, based on the ones I rated, and by finding patterns in other people's ratings. They're remarkably good at it. Yahoo has the same feature. Amazon makes purchase recommendations, but I find they're recommending things I've already bought from them, hardly very creative, and almost never results in a sale.

Anchors in OPML 

Anchors, as Tom Morris describes them, should survive a reorganization, and luckily it's not very hard to do.

Each headline that's to be linked to externally needs some sort of identifier that's unique within the file.

I suggest using the created attribute, which provides a unique ID down to the second. Of course this would fail if you create more than one node in a second, but I find that's a limit I can live with.