Directions Microformat -> Google for directions?

At Mix06, I had pressed Bill Gates a bit about the importance of microformats. In his reply, Gates mentioned that they could be useful for directions–and Tantek Celik, who was in the audience, picked up on this idea and made a call-to action requesting a microformat for directions.

Don Marti picked up on the significance of this idea: “When people put geographical directions up using microformats, someone will crawl them and string the route decisions together to get a directions search engine with common sense (because it borrowed the common sense of millions of users) that doesn’t tell people to make an illegal left into oncoming traffic, the way a certain map site used to tell me to leave my old house every day. (70mph combined speed motor vehicle slalom! Yaaaaahooooo!)”

This is exactly the train of thought that I’m sure Bill G was having, because in addition to our backstage discussion on the importance of microformats (which led to the exchange Tantek reported on stage), we also talked about how Bill thought that Navteq’s lock on directions data would come under attack as Web 2.0-style collaborative data gathering gained steam.

Tantek’s lazyweb call for a directions microformat as a way to start this bottom-up process seems really imporant to me. It would be really cool to see some progress on this idea by our Where 2.0 conference in June.

  • Realistically… I forsee the wisdom of the crowds concept falling on its face here. Quite frankly, if microformat crawling is the alternative to Navteq, then Navteq has absolutely nothing to worry about. However, a microformat for directions could be very useful for moving direction data around in an ambient sort of way. For instance, imagine if the Google Toolbar detected the directions microformat, then supplied a link that shows the directions superimposed on the relevant Google map or in Google Earth.

    But the crawling of directions and merging them together? Honestly, I’m surprised that anyone thinks that’s a good idea.

  • Bob — first off, your idea about microformats as a way of highlighting directions on a map is likely how this kind of thing would take off. It’s a kind of annotation. But I wouldn’t discount the idea of a collectively generated set of directions. I wouldn’t necessarily think of this as something generated by people directly, as by their devices. For example, as cars equipped with GPS follow their routes, the collected waypoints will tell us things like which routes are used most frequently, which way people never turn at an intersection (indicating a one-way street) and the like. Whether this data is actually expressed as a microformat is another matter.

    Probably the best place to start looking for how this might evolve is at some of the sites that help people to map their runs or other sport activities, like

  • Tim, that term “microformats” always threw me, because I was never sure what types of formats the speaker was thinking of, only that they were small.

    On the plane back to SF I found myself in line with Tantek (we hadn’t been face-to-face before, a surprise) and he mentioned that there were specific format conventions whose new support by MS excited him… something about particular “standard” formats. That put stuff in context for me, but I didn’t see the list of tiny formats, or an URL where they could be found.

    What’s the best way I can learn exactly what someone is thinking of when they spend a few minutes talking about “microformats”? tx.

  • John — go to

  • Tim, I already did, and read a bit, and it pointed me to a wiki, and I read a bit more, and I followed more links and read even more, and I’m still going “Well, who decides which is an *accepted* ‘microformat’ and which is an unaccepted one? Could you make something that you know would be an accepted as a ‘microformat’, or do you make something and hope that enough other significant people agree, or what? “

    (Anytime I spent lots of time reading about something yet still don’t feel I can functionally explain what it is to someone else, my BS filter starts to softly twing… not conclusive, just suggestive.)

  • Rohn Wood

    In an ecological sense, microformats could be seen as an emergent property of the Web ecology. Emergent properties are those which arise naturally from an environment but which are not generally predictable from observation. Emergent properties are also distinguished by their “fitness,” i.e., their robustness and suitability.

    Likewise, standards have proven themselves to be emergent properties. Likely there will many dead-ends, but eventually, the environment will weed-out the unfit efforts and the fit properties will emerge. But even these must adapt to a changing environment. The main thing is get the process started.

  • Tim, what happens when you have a small road running parallel to the highway? Many consumer GPS devices aren’t accurate enough to verify which road you’re actually on. That’s going to involve some really impressive algorithms to figure that data out, assuming it’s a realistic goal. And you still need to obtain the map data from somewhere to compare the GPS data against. And I’m going to go out on a limb and guess that if you’re trying to do the microformat thing, Navteq won’t be particularly happy with the idea of letting you use their maps for this purpose. As I understand it, there are open, volunteer-powered projects to create similar mapping data to what Navteq has, using similar techniques to what Navteq actually does. Closer to Wikipedia for maps and driving directions than Google. That strikes me as a more realistically viable way to go.

  • ‘Scuse me pressing on, but is this “microformats” label an attempt at another successful marketing coup, like “Ajax” and “Web 2.0” were, or is there an actual functioning definition somewhere, so that we can use it in a technical sense?

  • Jed

    I think the microformats discussion is a good one – as this the discussion on GeoRSS (see – but both of these deal with the idea of under the covers data compatability and hard things like calculating driving directions. But in this new era of location-aware/location-driven content (store/ATM finders, local search, shopping, even dating) we still aren’t defining the basic user interface. Every site with location content tags or identifies differently – some with lat/long, some with address, some with state, some with zip. But even within those uses, state is sometimes abbreviated and sometimes spelled out entirely. And don’t get me started on lat/long tags – even Flickr has a custom format for geo-tagging a photo.

    Why not a standard for web designers/developers on how they tag content and search for it on their sites? If we can agree on a basic standard, location-sensitive content can be shared cross publisher and easily retrieved with location determining technology (GPS, A-GPS, Wi-Fi systems, etc.) That means location-determining technologies can seamlessly identify points of interest, related user defined content and driving directions and present it in a unified way to the end user.

    Wouldn’t that deliver a better value to the end user?

  • Jed — I think you’re absolutely right about the babel out there…and that’s precisely the point of microformats, simple, bottom-up schemas that can be widely adopted.

    John — new names generally catch on when they mean something to a lot of people. Saying that they don’t have a meaning that you find comfortable doesn’t mean that they don’t exist. Read the introduction to Soros’ Open Society, where he talks about how people get hung up on what’s true and false, where many important classes of knowledge (markets, most notably, but also politics, social movements, and language) are “reflexive”, i.e. either true or false based on what people believe rather than what “is.”

    This isn’t to say that everything is relative. But it’s important to understand when new words frame concepts in a useful way, even if the underlying reality isn’t completely new.

  • pwb

    I’ve always wondered what the data looks like that drives the mapping services and to a lesser extent the directions services.

  • Man … thanks for the brain exercise, these comments got me going and my thoughts are below. My thoughts boil down to a system that stores accurate locations in the backend (lat, long, alt) and displays layers of applicable common names to the user.

    I am not a location expert, but I would think that simple, accurate locations would be defined by latitude, longitude, and altitude (in meters from sealevel). This is great for data generated by GPS systems which will include those data points.

    The trouble is that people names rarely include any latitudes or longitudes. So, unless our users regularly stumble across USGS brass locators, we need a way to translate common names to precise coordinates for storage and vice versa for display.

    Geocoding, is just this technique. It can take zip codes, neighborhoods, states, cities, etc and be useful in translating names to (X,Y,Z) coordinates. It is important to acknowledge the inaccuracy inherent in this because latitude, longitude, and altitude (LLA) records a precise place but names are rarely so precise. As long as we can compute an accuracy factor (A) to include in this simple format I think it will remain very useful. You can find a place to play with geocoding online at Ontok.

    The LLA for Carrboro, NC is (79.075272, 35.910000, 152.7048 ) which would put you exactly in the middle of town, probably not where the user was trying to tag. Use the accuracy measurement as a radius in meters, draw a circle around the point and assume and even probability that the tagged item is anywhere in that area.

    So, a tag of ‘Carrboro, NC’ might be (79.075272, 35.910000, 152.7048, 1500).

    If I gave the same tag as (79.075272, 35.910000, 152.7048, 15000 ), I might get Orange County or the Piedmont area of North Carolina, or possibly, both locations. This is where unkown power of a location microformat could take off.

    In translating from the precise tag, social tech can really add a lot of value allowing layers of colloquial names to come along with political boundaries, telling me that I’m in (‘Wicker Park’, ‘Sketchy Side of Wicker Park’, ‘Near the Blue Line’, ‘Chicago’, ‘Illinois’, ‘USA’)

    Anyhow, it seems like two things are needed: a microformat like the one I suggested (latitude, longitude, altitude, accuracy) and databases of location areas made of vector boundaries and end points in LLA form. These can be easily combined with powerful, informative results.