In spring of 2000 I was commuting every weekday, three or four hours each way by public transit -- nevermind how I got in such a silly situation. One especially crowded morning I started wondering how the net could help out.
Most obviously we could have a ride-matchmaker website -- you enter where you are and where you're going and when, and whether you have a car, and optionally other info like how many passengers you can take, how flexible you are about timing, smoking, etc... There can be reputation tracking like EBay has, to ward off knife-wielding maniacs -- plus big disclaimers to ward off suit-wielding lawyers. This could end up a lot more flexible than traditional ridesharing, not limited to regular commutes -- especially after mobile net access becomes common. And unlike most apps it would even get people meeting each other away from their computers.
A service like this needs critical mass to take off: no posts without searchers, no searchers without posts. We could grow to critical mass by providing some useful related service; my idea was unified trip planning over the different public transit methods in the Bay Area. (This probably would no longer work, as transitinfo.org now has such a service. It'd still make a worthwhile open-source project, though, and there are other metropolitan areas.) This would have some synergy with the rideshare trip-planning, too.
So I found some people interested in the project, sent out design emails, grabbed BART and Caltrain schedules off the web, and started coding the bare bones of a public-transit search. I got it working and complete enough to start prototyping a website around it, and that's where things stalled. I was worn out from the job and the commuting, and hoping the others would take over most of the work of building a web interface. But I've seen this sort of thing several times now: you can't expect people to contribute code until there's a working system that's very nearly usable for real, and furthermore it's as easy as falling off a log to write new code for it. I had no website, just the logic to go behind one: a Scheme program using real data to figure optimal routes. And the potential contributors weren't very familiar with Scheme. I was new to managing many-person projects; every time I try one of these I meet with new discouraging lessons. I hope someone else will take over this one, because it's still worthwhile and I have other fish to fry.
Here's what code there is. Write to me if you have any interest in taking it further -- there's more to the idea than is put down here. (Notably it needs geocoding using U.S. Census TIGER/Line files -- I can help with that for anyone serious about the project. I used to do that kind of thing at work.)
Phil Agre mentions a related idea among other applications of open geographical info systems:
Maybe mass transit based on buses with fixed routes could be replaced by something more like shuttles for airport passengers or people with disabilities: small buses that are dynamically routed based on people's geocoded indications of where they are and where they want to go.
Los Angeles now has a public-transit trip service at http://mtaweb5.mta.net/.
Thanks to Halsted Mencotti Bernard, Rob Duncan, Shae Erisson, and Jerry Hebert for productive discussions.