Good news, let's hope they spread it to Europe soon. While Techcrunch proudly post that 50+ million people bike in the US, that's only about 16%. Here in the Netherlands it's almost 100%.
UK-only currently, but seems much more advanced than the Google offering. It suggests three routes depending on your preference for speed or peacefulness and I like the little diagram that shows you how hilly the chosen route is.
This is great. Their walking paths work alright for this purpose, since there are many paths you can get routed over that are not roads. This project will add many new paths so that is a good thing. There are several issues with walking paths that this will remedy though - for example, you can be routed the wrong way down a one way street with walking directions, which I imagine will not be allowed in the cycling directions.
I just tried it with my old bike route to work. It actually did pick the route I used. In this case, the route is just the shortest path instead of the roads with the fastest driving time.
I also tried http://maps.cloudmade.com that someone recommended above. That one didn't pick my route. It picked the shortest driving time route. I guess this makes sense, since looking at http://opencyclemap.org, I don't see any bike maps listed near my house.
Incidentally, I just noticed that Google maps also has an option to get directions for public transportation, at least in my city. That's pretty cool. I knew there were buses running near me, but I hadn't ever taken the time to look them up.
This is exciting for our site, as our users want ways to see bike paths more clearly as well as better auto-routing along those bike paths. Driving directions crosses nasty highways, walking directions ends up routing over gravel and the like, so hopefully their biking directions is the 'just right' medium.
I am really surprised they didn't leverage the OSM data for bike routes. They used a group I had never heard of (though the group sounds pretty cool -- railways to bikepaths!). I wonder if they'll expand/revise using the OSM dataset.
I mapped a route from the Haight Street entrance of Golden Gate Park to the Castro Safeway. The route ends up being almost correct.
The route going to the Safeway bombs down Page street to Laguna and then wraps around. Overshooting like this sort of makes sense, but I would have just cut out at The Wiggle at Scott street and weaved over.
On the way back it used The Wiggle, but then goes up Page, which has a bike path, but is very steep. Elevation-wise, it makes more sense to do the full Wiggle over to Fell and then use the Panhandle bike trail.
Doh, you're right. I actually mapped nearly the same route (I live in upper Haight) and I saw it didn't use the Wiggle, but didn't realize it was because the route was downhill.
I wonder if it's based on special cases (like the mapped bike routes) or terrain.
It tends towards mapped bike routes. I don't think it would send anyone on Page for any reason other than it has a bike lane. I now live in NYC and it has mapped a few convoluted routes along the FDR bike path, when it would be far faster to use 2nd or 3rd avenue.
In some cities (Austin, Chicago, DC, Louisville, New York City, San Diego, Seattle) it's worth checking out http://ridethecity.com/. Not just because it lists "elevation gain" for each route (though I'm not sure if it chooses routes with the least gain), nor because it has "safe/safer/direct" option for routing, or even because once logged in it allows setting per-street routing preferences.
At a quick glance at Google vs Ride the City for NYC, Google seems to be lacking in a lot of on-the-street know-how (though of course has a fantastically better UI). Something similar happened with Google's public transit directions vs. http://hopstop.com/. Google seems to tend to idealize the model (assumes that intersections work, lights are timed, and trains run on time), while Ride the City and HopStop, for all their awkwardness, seem to understand the many things that can go wrong with seemingly clearcut routes.
