Craig Silverstein is talking about how Google Inc. works.

Some people think a company has to be evil to some degree to be successful, to play hardball. Craig’s glad that Google doesn’t think that way.

He’s going through general principles. E.g., “Good ideas come from everywhere.” (His example is the TouchGraph Google visual browser.) “Communications is key,” etc. It’s great to hear a company that’s walking the walk, but I’d rather have more of a drill down into the development process.

Ah, now he’s talking about the documents Google uses to organize engineering, e.g., their weekly engineering report. And they use blogs. (His example is evhead, Evan Williams’ blog.) He says no one speculated when they bought Blogger.com that they might use blogging internally. And now it’s working out great.

Google maintains a wiki-like page where people can contribute ideas for enhancements. They hold informal brainstorming sessions to evaluate the ideas.

They used to launch betas on the site and wait for feedback. Now they put it through some initial testing. They put experiments up at labs.google.com. (Craig particularly likes Google Sets. “It’s particularly useful if you can’t remember all the names of the seven dwarves.)

Craig says that Google is hiring. You can practically hear 500 people mentally composing their resumes.

(His slides are as simple as Google’s UI.)

I asked why Google only indexes 3 billion pages. Craig says that that’s how many they can crawl in a month.

Esther Dyson asks what they’ll be doing with their purchase of Applied Semantics, but Craig is amusingly evasive.

Question: The RIAA just sued a college because a local search engine discovered students were sharing MP3s. But Google is even better at finding MP3s. How worried are you?

Answer: We don’t let people search for music precisely for these legal reasons. We don’t want to make it easy to find illegal information. It’s hard to reconcile respect for intellectual property and make all the world’s information available.

Question: What are the barriers to all companies being like Google?

Answer: In part it’s because the enabling technology is new. Also, it takes a lot of work.

4 Comments »

L. F. (Felipe) Cabrera, Ph.D., is talking about Microsoft Web services. He’s trying too hard to convince the crowd that he’s a fellow geek even as he re-cycles a generic Microsoft slide set about Web services. (We know it’s recycled since he keeps telling us that we’re too smart to need this slide or that. )

Four design principles: Modular and composable; general purpose, standards-based and federated. Felipe elaborates on the importance of federation: “no central point of administration, control or failure.” “Federation forces you to respect all these different degrees of autonomy.” He’s stressing this obviously because Microsoft has been late to the federation party. Microsoft is focused on standards and interoperability, he says.

He’s begging us to interrupt him with questions. He thinks we’re “not awake.” [Or, perhaps he's just not being interesting enough.]

Now Tim O’Reilly is saying that the barrier to entry is too high because there is such a high stack of specifications. Felipe says that IP was once a technical hurdle. There’s a stack of specs because people want to do different things with distributed computing. Microsoft wants us to get to the point where the stack of specs is just taken for granted.

Judi Clark asks if Microsoft is now willing to work on open standards. “Absolutely,” says Felipe. “We hire more people to work on standards than lawyers,” he says to laughter and applause. But then he appends: “That’s Felipe’s perception.” And, he adds, “A clean design is better than a committee design.” [Ominous!]

Greg Elin asks if GPL is a guarantee of acces to code. What is the documented promise that what Microsoft says today will be valid in 2-3 years.

Felipe responds: “What’s GPL?” and totally loses the audience.

Question: Microsoft isn’t there with interoperability.

Answer: Yeah, it drives us crazy.

Question: In defense of MSFT, much of the problem comes from the XML Schema spec which is a nightmare.

Saman Far: How about having a reference implementation available in an open source way? Have you considered having the source available for the protocol stack implementations?

Answer: It has been hotly debated. I’m just a techie.

4 Comments »

Tough choice among sessions! A Microsoft researcher is talking about social software, Mitch Kapor is talking about Chandler, and Meg “Megnut” Hourihan is talking about “From the Margins of the Writable Web.” Meg’s always interesting and I love her title, so I’m here.

The tools for reading weblogs aren’t as developed as for writing them. Meg points to sites doing interesting things. E.g., weblogs that are tied to geographic areas. (You can put your geographic information into your blog via geourl.org.)

Also, sites are getting more explicit in their social relationships. E.g., create an OPML file of all your friends and put it in your weblog…

This is great stuff, but I can’t read the slides from in back and thus can’t get the URLs. I’ll get them from Meg’s site when she posts them. You should too. On to the Microsoft guy…

Marc Smith is talking about Netscan, a project for data mining newsgroups to see what we can learn about their social organization. For example, the number of cross-posted threads can indicate whether the newsgroup needs to fork. And 67% of Usenet threads have only two messages. Does this indicate success or failure? E.g., a customer support group wants short threads. How do you tell? One guy posted 95 times and every one was a reply. And posted 25 out of 26 days. He’s likely to be a high value “answer person.”

He thinks this type of analysis will be used by professional organizations trying to keep their discussion lists healthy. What makes an online community healthy? He says that you should look at things like time to reply, number of posts, percentage of messages replied to, retention of leaders, etc. He shows a user-friendly web page that Microsoft Research is trying to get Microsoft to build.

He demonstrates reading bar codes to get discussion threads about the bar-coded object. (It’s called AURA: Advanced User Research Application)

Fascinating. And Smith is a terrific presenter, getting laughter and applause along the way…tough for a Microsoft guy in this crowd.

4 Comments »

The Semantic Web, Ben says entertainingly, isn’t going to happen because the Web is too messy. So what does already have the data that’s required? A mailing list. Tons of metadata already there.

With threadsML (yeah!) you can capture threads. It’s “RSS 1.0 with extra toys.” It includes some Dublin Core metadata (metadata standard for documents). “Mod threading” lets you point to things that are children of the object.

This addresses the obsolescence of conversations.

The killer problem is: Subjects (topics). Why? Because subject lines are often not related to the topic. A solution is to adopt one of the topic hierarchies, e.g., Yahoo or Dmoz. But they’re culturally relative and brittle. An alternative is to create your own hierarchical ontology. E.g., Easy News Topics for RSS 2.0 (RSS 2.0 is RSS without the RDF components.) ENT was created by Matt Mower and Paolo Valdemarin. ENT points to a topic map and lets you specify a topic within the message. Matt and Paolo have today released an aggregator that reads the ENT topics. But Ben’s not sure how it’s going to work because even simple threads get long lists of topics.

What do you do about the proliferation of classification systems? When someone links A to B, the software can look up the topics that B declares, and it can assume that that’s a vote for saying that the topic of A is similar to the topic of B. [Interesting. This would work across languages.]

Marc Canter, who helped revive interest in threadsML, asks what are the benefits of having this sort of thing. Audience suggestions:

- Move threads from email to discussion boards to wiki, etc.

- Do it on one computer, moving between email systems, etc.

- Query it

- Handle not just text

- Link threads so the “backstory” is available

- Use email interface to read blogs, wikis, mailing list archives

- Connect reviews

(Disclosure: I’ve been involved with threadsML since the beginning, so I’m pretty stoked about this session. In fact, do a whois on threadsml.com… :)

15 Comments »

Clay is going to talk about a pattern in social software that consistently emerges: Groups are their own worst enemy.

Prior to the Internet we had lots of ways of doing point-to-point and one-to-many communication. Before the Internet, we didn’t have a good way to do “ridiculously easy group forming.” Now we do. But the patterns that emerge are due not to the technology but to how humans behave in groups.

Part 1: Why groups are their own worst enemy.

Wilfred Bion wrote in the middle of the 20th century. He believed that we are irreducibly social and individual. Bion found three patterns. First, groups engage in flirtatious sex talk among pairs, no matter what the ostensible group purpose. Second, they identify and villify enemies. Third, they nominate and venerate a hero beyond critique. All three of these are obvious on the Internet also. Groups “sandbag their sophistcated goals with these basic urges.”

Bion concluded that group structure is necessary to defend the group from itself. It exists to keep a group on track. Group structure defends a group from the actions of its own members. For example: CommuniTree, a BBS in the ’70s that failed when adolescents posted obscenities all over.

Groups come to a constitutional crisis where they not only need rules, they need rules about making rules.

Part 2: Why now?

If these things have been happening forever, why is it important now?

Observationally, there’s a revolution in social software happening. (Social software: software that supports groups.) Small groups are different than big groups. Now we’re getting weblogs and wikis and platform stuff that lets us try new things rapidly that support small to midsize groups, e.g., RSS.

Why now? Really: Why did it take so long? Why did it take so long to get weblogs, for example? WEW could have weblogs when we had the first forms-capable browser. Answer: It just took us that long to figure it out.

Second, this stuff is truly Web-native, unlike, say, Notes with a lightweight Web interface.

Third, we can easily put stuff together. E.g., Joi Ito’s conference call with a chat attached and then a wiki. It was a broadband conference call, but it’s a simple little thing.

Fourth, ubiquity. We can now assume, in many situations, all people have access to the Web. “All is a different kind of amount than most.” And within a meeting, we can begin to assume everyone is online. Clay no longer runs meetings without an online component (chat, wiki).

Part 3: Things core to social software

What makes a large, online group successful? After ten years of research, Clay can say with confidence: “It depends.” [Laughter]

The normal experience of social software is failure. Yahoo Groups, for example, exhibits a power law: few succeed. But there are about half a dozen things true of software that supports large and long-lived groups:

1. You can’t completely separate technical and social issues. E.g., you can’t separate the two mailing lists. “You can’t specify all social issues in technology.” The group can’t be programmed. Let the group decide what its value is and give them the tools to defend it; do not try to build the value into the software.

2. In a successful group, there is a core that cares about the integrity of the group and “gardens” it. The software ought to let the group express this fact. If you don’t, the group will invent its own ways.

3. Some members need more power than others. E.g., the Wikipedia’s “fire brigade” that undoes destructive changes.

What would you design for if you build social software?

1. Anonymity doesn’t work in group settings. I need to be able to associate who’s talking now with what’s been said before. Reputation is not portable from one situation to another: someone who cheats on his wife may not cheat on his taxes. Ebay’s linear metrics works well in a linear transaction system but not in non-linear conversation spaces. So, for social software to work, users have to identify themselves and there has to be a penalty for switching handles.

2. You need a way to recognize good members.

3. You need barriers to participation, some segmentation of capabilities. “Otherwise the core group won’t have the tools they need to defend themselves.” This flies in the face of ease of use, but focusing on ease of use looks at it from the individual, not the group. “The user of social software is the group, and ease of use ought to be for the group, not the user.”

4. You have to find a way spare the group from scale. “Scale kills conversations.” Metcalfe’s Law means the density of conversation falls off rapidly. You need “soft forking.” (LiveJournal does a great job of this.) Metafilter shuts off the “new user” page when they get too many users.

“The act of writing social software is more like the work of an economist or social scientist.” The people using the software will act as if they have rights. The site is theirs.

In response to a question: Old social software has been architectural: build a place where people can meet. Now we’re moving to a ship-building model: Build a place where people can go somewhere together.

[Great presentation.]

FUN QUOTE 1: “Learning from experience is the worst possible way of learning. It’s one up from remembering.” It’s better to learn from reading.

FUN QUOTE 2: Sophisticated computer graphical worlds look real “if you’re drunk and in the barrel” looking out the bung hole.

4 Comments »

Esther Dyson is blogging!

And so is Geoff Cohen, my town-mate and recently of the Center for Business Innovation.

3 Comments »

Kevin Lynch of Macromedia, a Platinum Sponsor of the conference, is showing an alpha compile of “Central.” It seems to make it easy to tie together apps. The first example is of a directory that pulls together info from a variety of sources. It looks a lot like a demo of a portal product but I’m sure it’s much more exciting than that. I just don’t know why. Maybe it’s the way the various apps (e.g., directory listing and the technorati link cosmos for the person’s blog) communicate and update. Via web services?

In another example, clicking on an entry in a directory pulls up an office map with the person’s sube marked. I’m not seeing what this is an interesting example of.

Developers can build Central apps and Central manages the distribution and fees.!

Aha! Tim O’Reilly asks the first question which is basically a statement about why Central matters: It’s like Sherlock on the Mac OS X in that it gives a new way of browsing. It combines the nearly ubiquitous Flash with Web services to treat the Web as a collection of re-usable apps and data, played in a rich client. Thanks, Tim!

Question: If you’re trying to replace the Web with Flash, what do you do about accessibility for the blind? And isn’t this bolted on top of an animation engine [and thus is inefficient]?

Answer: Yes, it is bolted on top. But performance is sufficient. And it runs across all the platforms on which Flash runs. As far as accessibility goes: Macromedia tries to be a leader, with accessibility support for Dreamweaver. Flash supports hooks into the accessibility features of the operating system.

Question: I feel like we’ve rediscovered client/server.

Question: Why not just do it in Javascript and SVG?

Answer: You could, but they weren’t designed for building apps, and with Center you don’t have to worry about a cross platform client.

Question: When are you going to improve the integrated development environment for Flash? We’re tired ot calculating pixel widths in order to resize windows.

Answer: We’re working on it.

[I guess I'm still bothered by the prospect of more and more of the Web being moved onto a proprietary platform.]

This just showed up from “Chris” on the discussion for this entry. I found it helpful:

It’s interesting because of the scale of the userbase (and it autoinstalls), and the fact that it will be everywhere – computers, cell phones, any connected appliance. One of the really neat features is the auto-updating, and working when not in Internet coverage. For cell phones, this is the most important aspect for apps, and one that each developer has to solve themself. Having a central mediator for all this potential traffic is useful.

6 Comments »

He’s going to complain and then show us some ideas.

He says the past twenty years of computing have been boring because it’s been focused on business. The predicted coupling of computers and the human brain hasn’t revolutionized human thought in general, even though it has happened in science. That’s because we generally are “instrumental reasoners,” i.e., we think about stuff only in relation to our current “goal structure.”

He shows a great video of the program Sketchpad which in 1963 was the first object-oriented graphics system: A room-sized half-MIPS computer that let you draw shapes on a screen. Totally cool. So early and so dead-on right: it is such a pristine example of the power of object-oriented graphics.

He shows a video from 1968 of Engelbart demoing what looks a hell of a lot like a personal computer. It includes communicating via video and audio with a remote worker. And another shows a 1968 graphical workflow designer. And, he asks, why don’t we have anything like Papert’s tools for teaching kids how to program? Kay is making his point that we haven’t really come as far as we think. We’re not making the breakthroughs we once did. He says it’s because “We’re not even thinking about such things.” [Or is it simply that the rate of innovation is always faster at the beginning? And perhaps we haven't done some of these things because there's no actual desire for them? E.g., it may seem obvious to Kay that kids want to and need to program, but he's an uber-technoid.

Noiw he’s showing a paint program for kids that looks a lot like an instance of Squeak programming environment. (Yup. I got something right!) It’s a slick and easy programming environment: to get the output of a steering wheel he just drew to steer the car he just drew, he drags a field from the wheel’s property sheet onto the directional field on the car’s property sheet. He gets applause.

Now he shows work by 11-12 year olds using the environment. Some elegant solutions to problems like getting a clownfish to find its meal by seeking the darker color gradient or keeping a car-graphic driving in the center of a twisting road-graphic.

Now he and Dave Smith are showing “pre-alpha” software that Kay calls a “broadband collaboration space.” Kay refers to Smith as “the slash in TCP/IP.” Given that the number of groups is 2ⁿ, how can we enable scalable group collaboration?

Answer: We’re looking at a shared 3D landscape. We’re seeing Dave’s view and Alan’s view on separate screens. The landscape is populated with framed photos of friends. You can look through a “portal” into another space.

There’s a cool demo of a flag waving in which the ripples are all dynamically computed. Likewise for an underwater world. And he draws a crappy 2D drawing of a fish and it gets turned into a nicely rendered 3D fish swimming in the shared environment. Kay draws a piece of seaweed. It too is rendered and shared.

Key point, he says: Late binding is good. He’s surprised that that’s not more widely embraced yet.

The audience gives Kay an extended round of applause.

3 Comments »

(BOF = Birds of a Feather, i.e., an open discussion with the chairs in a circle. This one is packed.)

Peter Kaminski has suggested a place where standards and ideas for social software can be proposed, kicked around, and maybe even implemented in rough and ready ways. The idea is to have a central coordination point to see if other people are thinking about the same thing. Goal: interoperability. He’s set up a wiki for it.

The meeting has gotten off to a bad start by too many people trying to decide how the email list ought to be forked; besides being a relatively trivial topic, the group has no way of resolving such issues. After about 10 minutes, Dave Sifry says that this is beside the point. The question is, he says,: Are we going to do something or not?

Pete asks how the group should be organized. General answer: Not. Let it emerge.

Discussion of the need for a broader understanding of the social effect of software. [Is social software necessarily democratic?]. Pete says that he can see a developer’s pledge coming out of this.

Greg makes an impassioned plea for seizing this remarkable moment in our history. And we should make the medium and small advances we can without waiting to figure out all the big issues.

Initial standards to discuss:

Kevin Marks suggests vote links: adding +1 or -1 to a hyperlink to indicate whether you agree with the object of the link.

Pete brings up Matt and Paolo’s topic standard, ENT.

Marc Canter brings up threadsML, near and dear to my heart. And an identity system.

Greg wants to manage images as collections of objects.

Saman Far: RFID

Dave Sifry: digital ID and stuff that connects the Web with “meatspace” (e.g., meetup.org). (Dave says he thinks we ought to start with email addresses as our digital identities.)

(I couldn’t hear some of the responses.)

A tentative meeting if only because most of the attendees are reluctant to impose order on an emerging social network, but worthwhile. Social software is going to be an important category.

Jason has a good comment on social software…

Be the first to comment »

Lisa Rein shows a brief video of the Bookmobile: a van with access to tons of books and a bunch of printers. Brewster Kahle says that the Library of Congress says that it costs $2.00 to check a book out and back in; he can print a copy of a book for less than that and charges $1/book.

Brewster says that traditionally we’ve been good about aggregating books in libraries. To preserve them, he’s digitizing and replicating. E.g., his group donated a scanner to Egypt and they’re scanning 2,000 pages a day.

There are about 16M books in the public domain in the Us libraries. There are about 8M from before 1923 and 8M from 1923-1963 that are in the public domain. About 20,000 have been digitized and accessible. With Creative Commons and Research Library Group, Brewster’s group is going to catalog the US reserach libraries to find out what we have and what’s out of copyright.

He talks six of his heroes, people who are putting themselves on line for the public domain:

- Rick Prelinger: digitizing film for free use
– Michael Hart: Project Gutenberg. 7,000 books keyed in so far.
– Gretchen Phillips: Digitizing children’s books
– Charles Franks: Distributed Proof Reading project
– The Million Books Project in India
– Tim O’Reilly: copyright of 14 yrs

Three things work against the public domain: Copyright, Access and License.

Copyright we know about: we’ve gone from a copyright of 28 years to life + 70.

Access: The public domain is locked up in private collections and libraries. It’s just hard to get ahold of the material. He warns that some companies that should know better are going to be announcing that they’re going to assert new rights over public domain materials, making it hard to spider them and access them. [Adobe?]

Corbis (owned by B. Gates) digitized a lot of photos from the National Archive and now claims that have the right to protect it via license. You can go an redigitize it, but getting access is difficult.

We need to bring public access to the public domain. It’ll take some work, some technology and some page-turning. It’d cost about $26M (a dollar a book) and the Library of Congress’ budget is about half a billion. At least let’s do the easy stuff: the public domain.

Terrific presentation. It’s great to hear someone who is doing good. And the book samples he passed around, fresh from the BookMobile press, are elegant and readable and cool.

1 Comment »