The weblog for the University of Southern California's (USC), Interactive Media Division (IMD) of the School of Cinema and Television (CNTV), looks like a winner. Continued regular participation from instructors and students on a number of interesting topics.
Putting some media back into New Media Hack.
Speaking of Rich Skrenta and Topix.Net, he also revealed some useful numbers regarding how many online news sources use RSS. Looks like there's still a case to be made for screen scraping if you're trying to be comprehensive.
Rich Skrenta provides a useful counterpoint to the LAMP dogma by exposing some of the thinking behind Topix.Net's implementation. In a nutshell, Skrenta says dump the RDBMS and stick to flatfiles. Be as dumb as possible.
I think part of the reason that the LAMP model is so pervasive is that querying data falls out of shoehorning yourself into the relational model. It's hard to design, implement, and optimize a data query language. Also, Skrenta discusses how to deal successfully with concurrent data access, which is another thing DBs like Oracle and MySQL buy you, although at the performance cost of complete generality.
Interesting stuff though and worth contrasting against Phill G's teachings. Maybe the flatfile guys need someone to write, or recommend, a book on the topic.
Well over a year and a half ago, Joe Gregorio made a connection between the Web and stigmergic systems. I've been thinking about del.icio.us as a potential example of the same principles: many individuals being completely selfish, coordinating through changes in a "physical" environment, to create a complex physical construct.
Looking closer at the definition of stigmergy, I think the analogy is not apt. The concepts of stigmergy have two assumptions that break down with Web systems. First, conceptually any individual in a stigmergic system can modify any bit of the environment. Clearly, everyone doesn't have the right to modify any web page. Second, in all of the biological and alife examples I've seen, there's been a Euclidean space that constrains the actors actions. As a network, if you use links as distance, you don't have a Euclidean space.
These issues may not be showstoppers, but if you want to model action on the Web as a stigmergic system, the translation is not a one-to-one mapping. It may not matter though, as this meme has seemed to die out. Probably better off going back to games on graphs as I first encountered them in Duncan Watts' work.
Tom Coates does some interesting interface thinking in regards to tagging and folksonomies. Most of the thought process is in how to incorporate tagged bookmarks into Safari, some of which has already been done for Firefor. Coates mainly approaches the topic from the angle of Flickr's tags but of course del.icio.us makes an appearance.
The comments are also worth reading, although of course a librarian, archdata, shows up and doesn't quite get the spirit of del.icio.us. As far as I read the zen of del.icio.us, tagging is purely personal, but just happens to be in public. The goal is not to come up with some beautiful, efficient taxonomy that everyone can share. That's too much work to build and to use. del.icio.us is designed to be as simple as possible, to get as many people to use it as often as possible. Splendid chaos ensues.
One sort of neat idea Coates had was the construction of ad hoc social connections being created through local aggregation of individual folksonomies. Imagine if Rendezvous was used to publish people's del.icio.us URLs. On the fly interesting intersections of tags and tagged urls could be built and examined. Don't really know where to go other than that the notion of on the fly, ephemeral, social media spaces seems like an interesting are of exploration.
Thought parking': wouldn't it be nice if there was a service that built a personalized, nicely laid out, collection of text and images culled from the blogosphere and dropped it off on your FooPod overnight? Audio's nice. Video doesn't float my boat. But something niche, tribal, and pretty for browsing in the interstices would be handy.
Lots of interesting stuff happening at John Battelle's Web 2.0 conference. I think I'll sit back and cherry pick a few events to comment on post hoc. But Jeremy Zawodny's overview of the Rojo demo lands squarely in my wheelhouse. I'm a big fan of social navigation applied to webfeed aggregation.
Scott Rafer, President of Feedster, in response to my podcasting post, chimed in to point out that FeedsterTV already exists. Thumbs up on the existence proof, but it illustrates my point about herky jerkiness as this stuff is figured out. The Feedster enclosure meta feeds are neat, but the search doesn't appear to actually be limited to media feeds.
Web services bustin' out all over. Niall Kennedy notes that Amazon has updated it's e-commerce APIs and made Alexa site information available. Hat tip to Greg Linden
Here's an interesting study, maybe. Take a set of "popular" URLs. Say the del.icio.us most popular, run them through Alexa's stats, ship off to Google's API, check out some other social bookmark services, maybe even see if Technorati has any recent information and see what you get. Perform on a regular basis and build up profiles based upon a number of different sources.
Call it 7 habits of highly successful URLs.
Way cool. David Temkin, CTO of Laszlo Systems announces that their toolbox of multimedia goodies has been open sourced.
All I know is that there are a few smart MFs that work there. Lisp hackers in the Paul Graham class. Macromedia and Microsoft may get some serious competition in that rich internet app space.
Addendum: Not to mention P. T. Withington. Anyone who's worked on real time GC for money is simply a badass, despite being Crimson. ;-)
I've really gone platform agnostic. Yesterday, on my office desktops, physical not virtual, I had a:
About December of last year, my Powerbook started giving up the ghost. Meanwhile, I was forced to buy a replacement laptop for one of my grad students, who subsequently didn't need it, ergo switch to Thinkpad as primary machine. In any event, I've been itching to get back to the Mac since NetNewsWire seems be busting out all over. And the malware situation on Windows is just really depressing.
Whelp, now I'm back in the Mac saddle. Best of all, I was able to dig up my license keys for NetNewsWire and LaunchBar relatively easily!! Yeehah!
While there are plenty of folks excited about podcasting, syndicating audio/video to portable media players e.g. the iPod, so far only Jeff Jarvis has succeeded in making the podcasting concept compelling to me. Partially it's his call to citizen's media, call it The People's Radio. It's also the acknowledgement that this stuff really isn't of the web. Sure podcasts are transmitted using Web technology. But their primary purpose, at least right now, is for offline content.
Hidden in Jarvis' post though is the reduction of these temporal media to URLs, the reuse of syndication technologies, and the potential for metadata. Basically you can pull all of the same stunts that are being done for URLs and text content. It'll take a while, but there'll be Feedsters, Waypaths, and Technoratis to take advantage of this ecology. Somewhere out there, a new Blogger and new SixApart are being cooked up (Webjay?) to make creating podcasts way easier. In the same way that Flickr seems to have wandered into a sweet spot for photos and the Web, the podcast groupies are fumbling about trying to figure out what works. Heck you could even imagine Topix.net and Findory style services. Eventually the big boys, Google, Yahoo, Amazon, will take notice and jump in.
For awhile it'll be a glitchy mess, concocted out of stuff we already know. Then the tools will start to peer into the media and things will get really interesting. I still think photos are the current frontier, but podcasting could be the next one.
Then again, I've been wrong before.
Robert Andrews documents the BBC's NewsWatcher, an internal Google News like service that automatically adds related links to news stories. This is a nice example of how a media organization can take advantage of syndicated content. While fraught with a number of hazards, e.g. playing up the competition, inappropriate links, etc., NewsWatcher style technology could help a news site add context without increasing manpower.
This could be especially useful if, as is increasingly true in the States, the news organization owns a bunch of properties that it would like to cross link. Now if Google makes Google News rebrandable, organization targetable, and privately web service accesible, they could make an offering to folks like Tribune Co., Knight-Ridder, NY Times, et. al. Don't know if they'd bite, but it could be another revenue stream.
Richard MacManus carefully examines his personal publishing channels and notices a certain proliferation. I think this will be the short term trend for such services, as the Web works out what it wants. There's still a lot of task focused authoring, publishing and aggregation innovation to be had. Once users vote with their feet and focus on a small core set of services, then you'll start to see some consolidation. But that's a ways off.
The one common thread that seems to be running across these services is webfeed formats as data exchange mechanisms. Other than personal notebooks, all of the tools MacManus describes easily generate webfeeds. Even the private stuff, other than paper notebooks, could be made available through password protected URLs. While current aggregators are good at slurping and displaying webfeeds, tools like atomflow are going to be essential to stitching all this production together.
Gary Price conducted an excellent interview with Findory's founder, Gregory Linden. Lots of good insights into where Findory started, where it's going, and the difference between personalization and customization. I'm a big fan of Linden's approach to machine assistance. Don't bother asking users to do anything. They won't do it, or they'll do it wrong, or they won't keep information fresh. The only hope is that they'll buy into letting you observe their behavior.
One point I'll quibble with is that "old news is no news". This is an issue I have with a lot of personal information management systems. Retrospection is higher order thinking that could be stimulated by many of these tools. To the extent that old news is around, is filterable, and can be summarized, looking at old news can put current news in context. I think this is also an element of "finding what you want", where what I want is some deeper insight and perspective on a topic.
Thinking out loud and not completely in tune with the article but hey... What if someone wrote a set of retrospection plugins for a popular blogging tool, say Movable Type, that just analyzed the blog archives and presented overviews and visualizations of the flow, warp, and woof of the writing. Or in the Findory case something similar using my reading history to give me a sense of what I've actually been tracking over time.
Marc Hedlund's article for the O'Reilly Network provides a nice discussion of using the new Bloglines web API. Many people are really excited about the feed caching, item management, and distributed synching that the services provides.
What really floats my boat though is the access to the subscription information. With a blogroll management UI outsourced, new services can be built that just look at people's blogrolls. If a significant number of people make their blogrolls public, then new interesting social services can be generated. Share Your OPML 2.0 could be just around the corner.
Also in yesterday's swirl of webfeed announcements, FeedBurner is supporting the usage of the Amazon Web Services API to splice contextual ads in feeds. This is of course why they're starting to think hard about licensing and splicensing as I termed it.
Link parkin': paramiko puts the SSH2 protocol in a Python module.
Could come in handy in a Python distributed system where you need to securely launch some processes on another machine.
Bloglines has been cooking up some Web APIs to provide access to its data. Two of the major desktop aggregators, FeedDemon and NetNewsWire , have signed on to take advantage of the capabilities.
In my earlier post about the new My Yahoo!, I was about to say "lookout Bloglines", but they seem to be zigging to Yahoo!'s zag.
The forward looking folks at FeedBurner are tackling the rather tricky legal aspects of splicing feeds together. This will be increasingly important as people try to integrate advertising with webfeeds. Note I'm ignorant/neutral on whether this is a good idea. The key nugget is that if you are seriously publishing using webfeeds, you better start marking your feeds and items with machine readable licenses pronto.
One of those issues I'm glad someone is thinking about, just as long as it ain't me.
Scott Gatz announces a revamped version of MyYahoo! which plays nicely with the open webfeed ecology. Jeremy Zawodny puts the right initial spin on who the product is targeted at, and why it's important. Think Hotmail for webfeeds and not just webfeed aggregation, but aggregation of all the personalized/localized content Yahoo! has its hands on. Not to mention a professional UI team to top it all off.
This webfeed aggregation stuff is for real.
Bertrand Sereno hijacks Flickr's photo annotation to create an annotated visual recipe.
Photos are the next first class editable element on the Web. Prepare accordingly.
Over at PassingNotes, David Carpe (sp?) has Spurl founder Hjalli Gislason's reaction to the acquisition of Furl by LookSmart. Spurl and Furl are in that same social bookmarking pool as del.icio.us.
There's a few interesting tidbits in there, especially that Yahoo! is looking to get in this game, and that all the search guys are looking to build new relevance ranking algorithms on top of this stuff. Google is called a wildcard, although I might point out that Google already has a bit of a social bookmarking service called Blogger.
For the longest time, I have been kvetching about the non-existence of extensible webfeed aggregators. Last week, while I was out of town, David Janes released J
So I find this to be really odd. At the WWW 2004 blogging ecosystem workshop, Cameron Marlow (blogdex) and Maciej Ceglowski (blogcensus) propose and present a new, open, blog indexing service, called upflux. Even though the service is vapor, there is zero mention of it in the blogosphere.
Nada, zip, zero. Or at least that's what Google, Yahoo!, Feedster, and Waypath say. I'd ask Technorati but they're down again.
At AIR 5.0, after Marc Smith's presentation on netscan, a service that's been archiving, indexing, and analyzing USENET for 5 years, that an open blogscan would be great. That's pretty much what upflux sounds like. Too bad it didn't get any traction. Especially since there's all these folks constantly complaining about how bad or evil Google and its ilk are.
Or maybe things are happening behind the scenes and I'm just fishing in the wrong ponds.
A while ago, I ran across the processing language/environment while examining the computational art of Jared Tarbell. I didn't really dig into the tool Tarbell was using though.
Ben Fry and Casey Reas' processing toolkit is centered around a domain specific language for doing interactive 2D and 3D graphics. The language is layered on top of Java so has a relatively portable development environment and can even be compiled into applets. As far as I can tell, there's nothing really interesting from a language design perspective, but the authors provide some comparisons with other languages.
The point of processing isn't to be radical though. Instead, the language is supposed to be quite accessible to those who aren't typical programmers. I don't know the background of all the artists that appear in the online exhibition, but whatever their hacking abilities, the works are pretty impressive. Besides Tarbell's stuff, the whizzy zipdecode application making the rounds of the blogosphere was done in processing
I've actually been thinking that a similar combination of domain language on top of Jython with prefuse inside would be really neat.
Link parkin': A nice little Python recipe, due to Christopher Prinos, for using Python iterators to fetch DB API results.
I picked this one up due to Simon Willison, although it seems to be kicking around del.icio.us as well.
P.S. Simon has a pretty neat trick with his blogmarks, click on one of the links and then revisit his page to see what I mean.
Apropos of nothing, I always get a lot of book reading done when I travel overseas. Not only is there a lot of plane, train, and tube deadtime, Internet access (used to be) spotty, so I was off the computer more than usual. (Actually, for this trip I will mention how amazed I am at how well Internet cafes work in the rest of the world. Not only are they ridiculously cost effective for users, they seem to be a very low cost and accessible way to get into business. They're the high tech version of the corner store.)
Anyhoo, I just finished the part of Steven Berlin Johnson's Interface Culture, regarding Vannevar Bush and the Memex. Johnson picks at one thing that really hasn't come to pass in the current version of the web, trailblazing, or constructing sequences of links and notes. If there's any a need for a new Web datatype, and accompanying authoring mechanism, trails would get my vote.
This seems like something that could be relatively easily cobbled together out of bookmarklets and blog/wiki software. Throw in some folskonomy, a touch of proactive search generated from trail construction, and a dash of social translucence to move us one step closer to the Memex.
Oh yeah. A nod to Johnson for nailing the blogging and journalism connection dead on the head. In 1997 no less.
If you're at AoIR 5.0 and wondering if this is the right Brian M. Dennis, yup you've got the right place.
If you're not at AoIR and wondering why I'm not blogging it, well one I actually like to pay attention in the sessions, and two the wireless access is a bit, uhm, lacking shall we say.
Arvind Satyanarayan and Elise Bauer have put together the accessible version of the details of dynamic publishing in MT. Combine with Brad Choate's overview, put in some reference elements (every detail of the Php framework) and you've probably got a halfway decent manual section.
Brad Choate digs into the dirty details of how Movable Type implements dynamic pages.
Just a thought, but it's a small step from there to doing your MT pages dynamically in Python. Ok. Maybe not that small, but it's an imaginable leap.
J
Speaking of plug-ins, there could be some serious innovation in the weblog authoring space if someone married a really nice authoring plug-in mechanism with a good publishing mechanism. For example, Movable Type has a nice, clean well-defined way to create and add new template tags. However, historically there's been no accessible way to do that for the admin interface.
Oh sure you can get in there and change the admin templates or even write your own subclass of MT::CMS but from experience, I can tell you this is not for the faint of heart. Application callbacks are coming along, but still bleeding edge.
This raises a general question about Web apps kicking in my head for a while. Plug-ins and lightweight extensibility are pillars of modern computing applications. Think Emacs, BBedit, Photoshop, Half-Life, Excel, Apache, Firefox, etc. etc. We know how to make desktop and server applications load code dynamically and interface with it. What's the analog for Web applications? How could developers, pro and amateur, write their own Yahoo! Suite, Gmail, Bloglines, TypePad etc. plug-ins? Does this even make sense?
Sounds like something to ask Paul Graham.
Gush 1.2 was recently released. (Did I mention that Gush is the most gorgeous IM app?). Mostly this release focused on PubSub integration, but the 2.0 features look mighty nice. The Conversation Gems feature, pulling out and highlighting things like phone numbers, sounds interesting.
Currently there are only two extremely minor nits I have about Gush. One, I can't get it to work on my Linux box (I admit I haven't tried very hard). Two, no plugin architecture.
Bill de h
A9, despite being an Amazon company, is still a bit under the radar. That'll change a bit after some recent pr blitzing. John Battelle was in the pack covering the new face of A9. He's got a little history with Udi Manber though so there's some added value above reports from the other usual suspects.
A commenter on John's blog, counterbalancing the enthusiasm in the piece, pointed out that a lot of the search UI knowledge has been around for a while. (Aside: for any tech article, story, report, paper it's a truism that "Foo did that 20 years ago!"). I think John's piece missed one key element in focusing on searchstream capture. A9 is trying to be of the Web. No highly interactive features. No fancy Flash or Java. Just text, links, styling, and some Javascript.
Another submerged aspect. Searches becoming first class objects and interfaces appropriate for managing them.
And a toolbar! More chrome to kit your browser with, at least on Windows.
Media companies should take very serious note of what A9 is doing. I would think advances in how people search for product might be of interest to them and their advertisers.
When I read Elise Bauer's description of the MTEntry plug-in, I mentally screamed "That's what I wanted!!" relative to my recent complaining about pointing to singleton entries in Movable Type.
Oh yeah, her discussion of ways to make MT entries sticky deftly covers many options. If you're really into modifying an MT installation, her site should be tops on your list.
Looks like a bunch of the INFOMINE project's tools released source about midsummer. As Soumen Chakrabarti advertised there is open source for an honest to gosh focused crawler: the Nalanda iVia Focused Crawler.
I'm starting a quick tire kick. It seems targeted at institutional libraries and is a touch bleeding edge. But hey, it's a potential starting point. Other than Filippo Menczer I don't know of anywhere else to get stuff like this.
Update: A January 2003 paper in D-Lib Magazine gives more context on the INFOMINE project.
We've been hack heavy here at New Media Hack, so to inject some new media into the conversation, check out Mike Mike's (not a typo) The Face of Tomorrow.
Take 100 digital headshots at one physical location. Get whoever you can. Composite, merge, transmogrify, those shots to create a statistical representation of people who pass through that space. Art ensues.
Low fruit extrapolations from New Media Hack: The Voice of Tomorrow and The Handwriting of Tomorrow.