Ramblings and Bragging: Convergence and Complexity

[gfish]

I finally broke down and bought a nice digital camera. (The upcoming honeymoon being the excuse -- and those in the legume gallery can just keep all their dirty thoughts to themselves.) I ended up buying the Canon Powershot S330, the cheapest-yet-solidest ultra-compact that had both optical zoom and an LCD viewfinder I could find.

This is the final major step for my life documentation infrastructure. I've been keeping fairly detailed text records in an emacs diary for a few years now. I already had a fairly advanced image archiving and indexing system, Erato, described here and here. I also have the VJ, a simply awesome system for playing video files onto the living room TV, described here. (It has greatly improved since that original post.) Up until now it was only for entertainment purposes, but the capacity of the camera to record short videos makes the VJ into a tool for life documentation as well. Some kernel building to get USB support for the camera, a few minutes of script hacking, and I can now seamlessly download images or movies from the camera and send them to Erato or the VJ, respectively. If I thought it would be useful, audio memos could also be downloaded and sent the DJ (like the VJ, but it has something approaching a personality and plays songs onto the stereo) as well. Now I just need to start scanning old paper documents and ALL OF MY LIFE will be preserved in perfect, non-degrading digital glory.

It's hard to see what the big deal about media convergence is. I've been working on it in my spare time for about 18 months, and I've got it pretty well down. I can't query the toaster status while browsing porn from my refrigerator, but I don't really need to, either. Nor is it generalized, or even slightly pretty, but it does what I need.

At the risk of falling for the mythical man-month trap, you'd think a big corp could throw a couple hundred developers at the problem and do the same thing. Of course, I used to work at a company trying to do just that, and they were terminally bogged down in an overly complicated, over-reaching, one-size-fits-none architecture. Not surprising; I always pictured building a home automation AI (SMAUG!) as a single, monolithic application. Instead I find one growing up around me as I fill niches one by one, with connections between them slowly forming. Rodney Brooks or Mark Tilden would be proud.

Comments

[shadowblue.livejournal.com]

*throws legumes*

[tithonium.livejournal.com]

1) You /know/ we can't keep our thoughts to ourselves, dirty or otherwise.

2) No, SMAUG wasn't monolithic, remember? System Management / Analytic Utility Gestalt.

[gfish.livejournal.com]

Backformed named aside, I never pictured the development process as anything other than 'we sit around and draw up a system architecture' as compared to 'we randomly write a bunch of little automation systems that somehow start to coalesce into a greater-than-parts-sum'.

Re:

[tithonium.livejournal.com]

I /always/ expected it would just be a bunch of little utils that do a specific thing, and all talk a common language that's put into a common pool of 'knowledge' about the state of the system, and they can react to it.

[gfish.livejournal.com]

The common language and sensor-fusion-y pools of knowledge would need to be designed top-down. But they could be applied to a pre-existing set of apps... Listings of data sinks, sources and transformations. Like how the DJ uses the TTS server on ibex, but dynamic.

Unfortunately, this is starting to sound a lot like Becomm bullshit. I'd need to check the non-compete to see if I can work on it yet, and even then I'd be pretty suspicious, since it sounds like Becomm bullshit.