Google I/O, In Retrospect

I went to Google I/O last week and thanks to DeWitt Clinton I gave a talk on where browsers are headed and if they can really get us where we want to go. I’m afraid that even more so than is usual for the talks I do with slides, this set is somewhat indecipherable without the actual talk along side it.

Slides are below in PDF format (6mb):

Sadly, I didn’t make it to many of the sessions, but I was able to catch Brad’s excellent talk on building a client-side searching utility with Gears. His demo app used a custom build of Dojo, which was exciting to see because not only was it stripping out stuff that he didn’t need from the base dojo.js using the customBase build option, but he was also able to use the build system to alias dojo.* to pt.* so that it didn’t conflict with other versions of Dojo on the page and also packed up all the modules into a single file. We’ve had each of these features in Dojo for a while, but seeing them used together was incredibly powerful. Dojo can act as “Dojo” or the basis for your own library without much more work. The demo itself (a client-side search engine) was also powerful. Brad used Gears’ worker threads to parallelize the work of pulling fetching, tokenizing, and handling a site’s content. The speedup in being able to move this kind of work into the background (potentially onto separate processors and cores) opens up a new world of potential applications. I’ve been thinking about the implications of a ubiquitous Gears ever since.

An aside: Google throws a hell of a party. The only thing I’ve seen comparable to it was the awesome time that Microsoft hosted at this year’s MIX. Landing the Flight of The Conchords was quite the geek coup for I/O, particularly since most of us had no idea who would be playing. Considering that I also got to see David Capurro at yesterday’s Laughing Squid party, I’m pretty blissed out on geek meme entertainment.

Update: video of my talk is up at the Google I/O site. The video should make the arguments somewhat more clear than the slides alone could (although on the downside you’ll to suffer through my myriad “um”‘s and “uhhh”‘s). Also, Neil McAllister has a piece up today which summarizes some of the talk’s points. I’m afraid the talk left him with the impression that I support natural monopolies when in fact I only raise questions about their formation in order to find ways to effectively break them (or hollow them out).

One Comment

  1. Posted June 4, 2008 at 8:18 am | Permalink

    Hi Alex! Thanks for the nice feedback. BTW, I renamespaced Dojo to ‘pu’ instead of ‘pt’. I’ve also backed off mostly from making a custom Base build. At first I stripped out Dojo Query and the base FX manipulations. However, I couldn’t keep myself from using Dojo Query since its just so useful :) I could knock things down to about 32K minified non-gzipped, but the fact that Dojo Base is just so small gzipped means that making smaller slices of it doesn’t necessarily matter for this project. However, being able to do this cleanly in the future is really nice using the customBase property; for example, I sometimes have projects that just need the Base language utilities, package handling, and basic XHR support, and being able to use customBase for these is really nice. The customBase and renamespacing stuff (along with being able to use D.O.H. for external projects in a much easier way, which I’m also using) are all really nice new features.


One Trackback

  1. […] about Dojo, Peter is now your go-to guy. I’ve enjoyed having the opportunity to think and talk about where the Open Web is headed, so I’ll also be doing a lot more of that. There are lots of meta-issues that this transition […]