Outsourcing My Stuff

Life has been, as usual, hectic. Since we’re renovating the flat, I’ve had to make a few adjustments to what I can only loosely term “the household infrastructure”, which consists of a ragtag assortment of computing and network hardware that would not be out of place on your average computer lab.

There are, at the last count, four small Fast “Ethernet”:Wikipedia:Ethernet switches, three active routers and seven machines strewn about the house, ranging from my to game consoles – and their being off most of the time does not disqualify them from being networked in the “old school” way (i.e., with Cat. 5 cable). The bandwidth I have available is… interesting, and I would term it wasteful if it wasn’t: a) part of my work and b) temporary and subject to constant revision.

But yeah, I need to simplify things a bit. The bottom line is that the house server (my trusted G4 ) will be down for a good while as walls go down near the place it nestles in, so after some consideration I decided that, given that I do less and less stuff on my machines and spend most of my time using for work anyway, there is no reason whatsoever to avoid trying out this brave new world of web applications and services.

That and a nagging feeling that the way I use computers is shifting profoundly and that (despite my marked preference for decent desktop environments like the ) I need to adapt to a point where I don’t really care what I’m using, so I’ve been moving archives and stuff from my home IMAP accounts to (for news clippings, non-critical personal e-mail and other stuff I only really need when on a ) and Apps For Your Domain (which is where most of my mailing-lists, site-related mail and whatnot have been migrating to).

And I’m going to try doing it all the way – for instance, this post is being drafted in Docs.

Which now mostly works in 3.0, although the resulting HTML is anything but palatable (in fact, it is messy as anything) and I end up running it through detextile.py (which I have installed as a Services menu item on my , courtesy of ).

But, more importantly than that, I’ve started migrating my RSS feeds to .

This is not a wholesale endorsement, but the tipping point came when I realized that it now works more or less OK with the mobile devices I use (although it could do with a major revamp for, say, the and the , where the layouts completely suck…).

It’s an uphill process, however, since their UI still needs a lot of work (it isn’t that pretty, not does it make very efficient use of the top of the browser window) and does not do three things my current setup delivers flawlessly:

  1. Decent archiving (I receive RSS items as a MIME-compliant mail message, with all the images included in the message, not just references to them)
  2. Bayesian classification and filtering (i.e., stuff I’m not interested in never makes it to my screen, which is delightful)
  3. Screen scraping (for those pesky feeds that only have headlines, runs helper scripts that grab the rest)

So I’m now basically making do with a heavily customized set of Yahoo Pipes to aggregate and try to filter out most of the junk, as well as trying to figure out a way to use shared items as part of the feedback loop for Bayesian classification (by tagging stuff I don’t want to read and “sharing” it back to a classifier, which will act as a proxy for some RSS feeds).

But that’s a pain in several ways, since the tagging mechanism isn’t one click, and having to deal with different feeds for positive and negative classification is a bore (not to mention that my running the feeds through a classifier proxy creates an additional point of failure, and Yahoo Pipes is a bit slow as it is).

I don’t think there’s much chance of adding wholesale Bayesian classification to , but decent filtering would be very nice indeed – I’m getting tired of having to click through oodles of rumors as MWSF approaches…

This page is referenced in: