Laughing Meme

One of the key topics (I think) in my Casual Privacy talk last week was the importance of “context” in privacy and sharing. That some people have trouble understanding how fundamental context is to all social interactions was my primary take away from SG Foo, and I’ve been preaching it quietly where I can.

All by way of saying, I made one of my rare visits to FriendFeed this evening, and I was reminded that I consistently regret it. Breaking down those contextual walls means I consistently like the people I find there less then I did when I was able to interact with them in isolated manners; fire walling the aesthetic from the technical from the political from the personal.

We need routing not aggregation.

Learned a new word tonight from MattB, SimonB, and Yoz.

A hashbot is a robot that hangs out on an IRC channel (hence the #) and provides a conversational interface to a resource.

hashbots are the ancestors of Social Software for Robots, and the idea of Twitter/YubNub as Web CLI.

You might have seen the post on the Flickr blog announcing Places, or maybe the Good Reverend’s write up, but if you haven’t:

Places is a new Flickr feature that mines our corpus of geotagged photos, identifies characteristic features on a per location basis, and then goes back into the data looking for “iconic” beautiful photos. (btw try reloading that /places page, the feature places are random. As to a certain degree are the photos on the individual Places pages themselves)

It also is where a good chunk of my creative energy went for the last few months which is why the blog has been so quiet. And its a hell of a lot of fun, not to mention a privilege and pleasure to deep dive into our database and be reminded just how much fabulous photography there is on Flickr, and maybe just barely fumble around the edges of surfacing the diverse communities shared vision. Eyes of the world indeed.

A Place for GeoRSS feeds

Dan roped me in on Places months ago. We had geoFeeds working for semi-arbitrary places, and we needed a page to hang them off of. That page looked a lot like search result. You never saw it because the Flickr project management process (a blog post of its own) left that particular prototype a bloody, heaving wreck. Don’t worry, the current version is much much much better. (of course you also never saw Dan’s brilliant prototype of the current version, which was too cool to release on an unsuspecting public) And voila, many months later, the feeds are there. (though I’d still like to bring back that SRP view to allow rich searching within a location)

Increased Surface Area

We brought a bunch of different design goals to Places, but one of my obsessions that I think we nailed was the idea of “increasing the surface area” of Flickr. (also known as providing new ways to level up in the Game of Flickr[tm]). Only a few people, and a limited range of styles will ever be featured on the Flickr Explore pages. Which is fine, most people don’t care. But Places provides another way to recognize the contributions of Flickr members, by hilighting their geotagging and their photography skills. I’m looking forward to adding a couple more similar features to Places, recognizing other Flickr Games one can level up in, and other contributions back to the commons you can make.

Mo’ Betta

A bunch of stuff didn’t make our initial launch. Some of that has come in since then. More will be coming. I’m particularly excited about using adding some new data sources to improve the page. (e.g. the Groups right now a bit weak, and we don’t have reliable neighborhoods in cities, both of which are in process of being fixed)

Thats kH8dLOubBZRvX_YZ to You

Turns out there are a lot of San Franciscos in the world, and we personally struggle to keep track of which one is which. So we’ve been experimenting with giving them unique place_ids. If you look really close you’ll start to see these popping up around flickr, in photos.getInfo, photos.search, and as microformats on the Places pages. Its all very experimental, this unique identifiers thing, but we think it might work.

Arm Chair Travel

And because I love you, I’m going to let you in a on a secret. Have a great trip.

Taking a break from all the collaborating, originally uploaded by George.

I didn’t make it to FOO this year, but I did send software in my stead, and its nice to hear that folks liked it.

We slaved iCalico to Crowdvine to add a social networking layer, a network that was walked, mapped, and color coded by the Pathable folks.

Tony has a nice report back on it, as does Shelly from Pathable (6 weeks aka a couple of late nights). And Scott Berkun (who owes me a copy of “Art of Project Management”!) said super nice things.

Collusion Patterns

So how do you do that — stitch together 3 different sites to provide a unified experience? Visions of APIs, Internet scale SSO, and messaging layers spring to mind. Or more likely hash and slash patches, jury rigged shunts, juggled install directories.

We did the dumb easy thing, and I’m surprised more people don’t do it.

1. Crowdvine.com sets a cookie collusion. This cookie contains the data we needed to display the logged in view of iCalico. (you’re nickname and optional your URL). In addition it contained a md5 hash of the concatted data, plus sekret known only to Tony and myself.

2. If we find the cookie collusion, we load the described user from the database, or create it on the fly behind the scenes.

3. There is no step 3.

Amazingly useful, trivially simple, ultimately flexible. Niche sites are great, but you need techniques for stitching them together before they can realize their potential as pieces of an ecosystem. I don’t necessarily expect to see this kind of integration become more common, but I think it would be great if it did. (and in the name of transparency disposable apps are huge enablers, disposable sites/apps is another pattern I’m puzzled we don’t see more of — its as if we more inclined to converse bits then landfill)

update: Whoops, it was pointed out there was a step 3, or rather a step 1.5: use CNAMEs to point to individual components on sub-domains.

NYTimes has an article on the ongoing Netflix recommendation open challenge thing. ($1mil to a team that can produce the best collaborative filtering mouse trap)

Unfortunately the project is flawed, because the basic question is flawed, fundamentally and in a very simple way. We have moods, we have shifting interests, and trying to compile all those multi-variates into a single vector of interest is impossible.

Rather then making the computers super smart, I’d rather see an interface like the Pandora channel creation where you choose 2-4 songs that suit your mood and the system finds the common elements.

Tonight at the video store I wanted something that was smart and fast enough to be engaging, without being so smart that it took work to follow. Maybe a political thriller? If I could have mixed a recommendation queue out of 3 Days of the Condor meets Wag the Dog meets something like Enemy of the State to find something in that vein that would have been better then all the weighted neural nets.

And I’ll wave the million dollars if they just build it already.

Not sure where this systemic biasis for computer as deep thinker comes from, probably dates all the way back to the Ultra project and other primordial computer science legends. But its the wrong metaphor here and now, smarter, smaller tools to extend the human reach, not replace humans.

(can you tell I didn’t find my movie?)

Stone Cottage pointed to a great post by Rick Skrenta, CEO of Topix (and mad mind behind NewHoo/DMOZ for those who can remember back that far) on the Topix re-launch.

Lot of really interesting stuff about identifying a brand’s core value and putting it into practice. But also a phenomenal laundry for a problem that has stumped a lot of us, how to make a local news site succeed. Including:

• Anthropomorphize our existing technology into the roboblogger. This was a brilliant idea from one of our lead engineers. It simultaneously solves three problems:

  1. Booting up a new city — you need posting activity to draw the first editors. The roboblogger would give us that. But he is shy and gets out of the way if humans show up and take over a page.
  2. If the community editors go on vacation, the roboblogger can step back in and take over while they’re gone.
  3. People know when a robot is editing the page vs. a human. His profile icon is a picture of a little tin-can robot. His handle is ‘roboblogger’.
  No more confusion.

A project has to already have value to draw a valuable volunteer base, this is the classic and yet fundamentally hard problem with boot strapping all local sites. But as soon as you have volunteers your contract with them is to rain attention and love down upon their contributions. roboblogger is a really neat hack to handle the delicate balance of a site’s lifecycle and mix community and data mining techniques in social software. Looking forward to watching it play out.

On a list I’m on there was a proposal to do a quick salary survey among the members. The proposed solution for handling this was

Everyone say ‘I’m in’ to get onto a list of N people participating. Then write down N numbers, positive or negative, that add up to your salary and send 1 of them to everyone else on the list. Add the N-1 numbers you receive to the number you didn’t pass out, and announce that to the list. The sum of those numbers is the total salary. If N-2 people gang up on you they can’t guess yours. As a bonus, you can always back out at the last minute before divulging your own sum without having given anything away.

Which you got to admit is a beautifully simple, and elegant solution to the problem of sharing sensitive information in a low trust environment.

The solution? A 7 line Perl CGI to just write salaries to a log, and a good faith promise not to track IP addresses.

I missed it, I’d even seen the inimitable Wilfredo on the lists and it went right by me, but I ran into Nat last night at Buzz’s WWDC party, and he rather riveted my attention (as I think he knew he would), that iCalServer announced yesterday? Turns out its an open source CalDav server, written in Python (Twisted), Apache 2.0 license, with great unit test coverage. (which bodes well for the trial by fire known as interop)

Btw. Collaboration’s homepage seems to be a Trac install which is sputtering, and crying. Under load?

update: Um, and how long as Cyrus Daboo been signing his emails “Apple Software Engineer”? Yup, I was asleep.

update 2: wsz confirms trac unhappiness (and my Twisted inside observation), I guess I just got in in time, if you email me, I’ll send you a tarball of the src.

Ah yeah! Haven’t played with it in depth yet, but this is the service I’ve wanted forever (and Carl and I seriously sat down and threatened to build).

From Joi Ito

It reminds me a bit of BookCrossing, but the approach is different. BookMooch is more systematic. On BookMooch, you register your books and others search for books. You use points to get books that you earn by listing books. Unlike many other used book services, they don’t get involved in the shipping and payment. It’s rather peer-to-peer.

Which is all I know so far, but I’m so diving into this tonight. From the creator of Magnatune. (I wonder if they’ve already got selective importing from the various book listing services

Private, friends|family, do not share, etc. It’s very useful to be able to flag certain data you store as private; links or photos, event attendance, or notes. These simple defaults allow us to manage our data without the complex overhead of maintaining ACLs (a task almost no one is obsessive enough to manage)

But it would be nice if there was a way to re-define privacy on these sites so as to enable sharing among partners without the single account/password swapping solutions. Looking at the statistics I have to think this is not a niche feature.