It was also an experiment in extracting structured data from semi-structured, free form wiki like data entry.

I moved away, and the site kept running itself. Eventually the spammers overwhelmed the community and I had to shut it down. I feel bad about this. Bringing back (and rewriting) WifiMug is on the todo list, it’s on the given-the-ability-to-freeze-time todo list. (by far the longest of my todo lists)

So I spent 15 minutes attempting to do the right thing, and all the data (and all the spam) is available for download under a Creative Commons Attribution-ShareAlike license.

The dump contains the following directories per city:

• database contains the raw text of the wiki pages that represented each of the cafes (and the various meta pages).
• metadata contains some of the structured data, per page
• rcs the history of each page. (yes rcs)

Ideally I’d do more. Ideally I’d scrub the data, port it to some sane format, tease out the implicit metadata encoded in the markup, attribute all of the various community members, etc. But I’m trying to not let the perfect be the enemy of the ok. This is a minimal competence thing.

And ideally we’d should be able to hope and expect that Yahoo! would do something like this with the infinitely more important and influential del.icio.us data set — dedicate a few weeks or months of time to preserving one of the greatest new libraries of our time, possibly donating it to Archive.org or the LoC. But nothing makes you uncomfortable like holding Y! to a standard you aren’t personally living up to.

Get the data.

All done/written under the influence of 30k feet and 10 minutes of reflection, treat accordingly.

ps. I almost got all of this into 140 characters, but failed. I hate the way blog posts feel so flabby and fluffy after the compressed kinetic energy of a tweet. I mean Anil makes me feel all noble doing it, but I miss my creative restrictions.

Enough frisbee, need to get a little work done this weekend. Need a good coffee shop with wifi and power on a Sunday.

Hmmm, really what I need to do is to finally get around to setting up a San Francisco WifiMug.

Which reminds me that I’m kind of unhappy with the current state of the WifiMug codebase (and it’s showing its age a bit, pre-Web 2.0!), guess I should research available Rails wiki software.

Though what I really should do is whip out that Wiki-meets-structured-data-creation-PlacesWiki I’ve been meaning to write.

But first I want to do a comprehensive survey of alternative wiki markups, the space has really fractured. Stikkit is doing good things with natural mark up, while Semantic MediaWiki steamrolls its way through the expressiveness sweet spot, and Google Code wiki has some interesting stuff going on.

But, um, um, um, I should write a new web browser and search engine to research wikis! Yeah, thats the ticket!

http://www.idealog.us/2005/07/laptop_discrimi.html

CafeSpot, a social guide to independent cafes, coffee shops, restaurants and more.

So the obvious question is, how does this project benefit from and share data with other projects like Delocator or WifiMug?

Someone should really get together and coordinate a data interchange format/protocol/consensus.

http://akuaku.org/archives/2005/04/google_maps_on.shtml

go wired on weekends. It would be sort of retro, you know, in keeping with Victrola’s theme. Put an eight port hub in the middle of the coffee table in the back, and make all of those antisocial net junkies (like me) sit next to each other on the couch…

During my sparse visits to UZ I noticed that an unofficial “laptop table” existed, with the Monster kids as its core. It had the interesting effect of setting up a second social scene. No one who has brought their laptop to a coffee shop to work will be a quality participant in the discursive, casual cafe scene, but that doesn’t mean one can’t create another equally vibrant scene built around our shared work. (Useful in particular as most of us are working in fundamentally alienating profession anyway)

http://www.sims.berkeley.edu/~savage/ps/

What Is It?

In brief it is a website that attempts to offer a national (presumably U.S. national as it asks for zip codes) database of independent coffee shops, in order to support those fine and public places.

Things it gets right (Web 2.0-ish):

• geographic – the physical was missing from the web for so long, that even though it is becoming standard each new website which can tie the virtual experience to real world location presents a little epiphany.
• user built – this is an art project so the mail off a stack of phone books to China and pay to have them typed in route was never an option, but still its clueful to see someone make user contributions front and center.
• its pretty (ok, that might not be Web 2.0-ish)
• its opensource

Things it gets wrong (Web 1.0-sih):

• roach motel – why should I contribute to this database when there is no way to get a dump of the data?
• no maps
• no user editing – a quick scan of my area shows it is not only sparsely populated, but that of the 3 entries I did find I could add additional info to all of them.
• no user profile, no community, no reputations – you can see that I added June Bug, but rather then my name being a link to my profile, its a mailto with my address!
• no permalinks – can’t really expose a database of first class web objects without permalinks
• a splash screen, with a popup window!?!?! I feel like I’m in a timewarp!

And just to be clear, I’m not attacking Delocator for this, my own minimal attempt at cataloging and promoting independent coffee shops falls down on most of of these points as well. Just talking them through.

A Short Story About Roach Motels

So why would you possibly want to provide a dump of your entire database? Re-use and re-mixing. Projects like delocator, openguides, addyourown, et al. are one facet of how we’re starting to annotate our spaces around us. Projects like mappr are another. THe more we can get the data out of it’s silos, the more we can combine it to interesting effect. (and if we can just get it all into RDF we can sit back and let Jo do the rest)

But a simpler story is, when do you want info like Delocator (or any of these) provide? When you’re out. Not when you’re sitting at home in front of the computer. Opening up your data means you can get someone to help you with a mobility solution, be that a cell phone based interface, an iPod compatible database, or a clever PDF to print out and stick in your pocket.

Two Way Data Interchange

What we really need is a data format for this stuff. I personally I know the website, address, phone number, etc of about 100 independent coffee shops not listed on the Delocator page. (call me obsessed) And I have most of that information stored digitally. If I had a way to send them an XML file of that information we’d both be happier. Similarly I’d be happy to contribute to addyourown, chefmoz, and openguides, and would love to be pulling out the data from those sites to enrich my own listings. But not if I have to re-type it!

State of the Art

I did a brief survey of available formats last Summer, and didn’t come up with anything compelling. The ChefMoz format looked like it might be a decent starting point, I no longer remember what I found so problematic about it. Anyone else interested?

update: a bit more on Delocator

From 3 locations to 7 in a couple of hours is pretty good growth, I’m impressed. And I wouldn’t have thought to add City Feed, which is one of my all time favorite places. (just decided what I’m doing for lunch!) But how do we define non-corporate? In Boston this is particuarily hard where almost everything is part of a mini-chain. I’ll grant you Emack and Bolio’s with its 7 locations in Boston probably makes the cut, but how about ERC, with its 20 locations in 6 states?

For example Lattice’s Hacking Google Maps for Caffeine and Wireless, inserts custom data into the Google Maps data stream to produce zoom-able, pan-able maps of Vancouver, Seattle, and Boston enriched with information about the locations of wireless coffee shops. We’ve only begun to explore the surface of what sort of metadata you can insert into the stream, and what other services you could build on top it, but the potential is just a little mind blowing. (actually to hell with the potential, I find the current incarnation a little mind blowing as well)

Go for it, take a virtual caffeine tour of Vancouver, marvel at the Capitol Hill wireless coffee shop scene which is rightfully the 8th wonder of the world, zoom way way out to survey our Boston options

One of the big question marks is Google’s support for this, as they could break it with an eyeblink. (as they did with simple XML output from Google Maps) Lattice has an interesting point on that:

Now, considering that all requests to Google are being performed by the users’ browser, and I’m just injecting some JavaScript, it smells a lot like AutoLink

I’m not sure how well that argument would fly at corporate head quarters, but I would suggest that there is run for a serious long wake play, and they should ponder long and hard before moving.

http://boston.wifimug.org/

http://seattle.wifimug.org/index.cgi?MCafe

Interestingly since it ran, WifiMug has been overwhelmed with people whom didn’t understand the wiki, and kept deleting the home page, (only 1 case of intentional defacement, and that was after the front page was already a wreck), plus several of the mentioned companies adding inappropriately commercial pitches.

I finally locked the front page, and need to think a bit more closely about how to solicit contributions (which until now have flowed in smoothly)

Still its always fun to see yourself in print, and this was no exception. Thanks to Jon, and Jeff who spotted it first.

update: Wow, no one told me it was on the front page! Also I’ve unlocked the front page of wifimug, lets see if it survives this time.

http://seattle.wifimug.org/index.cgi?SeattlePublicLibraryCentralBranch

The wireless coffee shops of Seattle wiki is now at its own domain, wifimug.org!

The best part about moving the wiki to its own domain (besides the boost in credibility) is this opens up the possibility for having a whole federated network, of loosely autonomous locale based wikis of wireless coffee shops. (in other words, write me if you want a subdomain for, say boston.wifimug.org, or montevideo.wifimug.org)

Random Wiki Thoughts

• Squid and wikis do not work well together. Its amazing how confused people can get when their changes don’t show up.
• Wiki’s generate very little Google power, especially given their inherently linky, twisty, wordy nature. I blame WikiWords, and think perhaps wikis would be better served by displaying the above as Wiki Words.
• RSS feeds of RecentChanges is match made in heaven.

Next Steps

• Wiki in a box – it is easy to tar up, and download a Kwiki wiki sans version history (RCS gets grouchy about changing file permissions), making it easy and obvious to keep a local copy of the wiki on your laptop might be cool.
• Auto-generate a PDF with names, addresses, and other key pieces of information (maybe on a personalized basis) for easy print out, and stuff in the pocket.
• Palm sync/iSync support