June 19th, 2002

I’ve been having a look at Amphetadesk today.

In conjunction with Les Orchard’s collapsable-channels-and-items skin, I like it quite a bit.

I originally loaded it on Win2000 and browsed to it on the same machine as intended, but decided I wanted to set it up more like my usual configuration.

My current modus operandi for news aggregation is to have Radio always running on my home machine and then I connect to it from wherever I happen to be to get the news page. I can get it from the kitchen, from the office, from someone else’s desk, anywhere.

So, I loaded Amphetadesk up on my Linux box and ran it nohup (that is to say as a constant background service).

I imported the same 39 RSS feeds I use in Radio by simply taking the gems/mySubscriptions.opml file from the Radio directory and overwriting data/myChannels.opml in the Amphetadesk directory and restarting Amphetadesk.

Everything loaded up nicely – 39 feeds, each with multiple items, all nicely collapsed or expanded at my behest.

Worked fine from my Win box at home too – took a while to load the page, but not too bad.

I told Tim via IM to come take a look, and it took him AGES to load. I looked, and of course, since every RSS feed was being not only loaded but also rendered to verbose HTML, the page was over 1.1 Megabytes.

Even after dropping down to only 6 feeds, the page still loaded up at 100k. You can see this won’t scale well at this rate – if I were to load up Jenny’s feeds, I’d be clocking 2Meg of dload per page view. Ouch.

Another issue for me is that feeds are listed by latest downloaded. From what I can see, they’re downloaded every time there’s a scan though, so every feed always seems to have been downloaded at the last scan. I’d rather have their RSS content compared upon download to the existing content and then have the list sorted by last changed rather than last scanned – this would make the recent data percolate to the top, which it currently doesn’t do.

Then one could highlight the recently changed channels, populate them with content, and only load or view the content of unchanged channels on drilldown, either to another page or dynamically filling a div via xml-rpc or remote scripting. This would help the page to not be so big until needed.

Overall, it’s a great hunk o’ code – really nicely written and documented. In an interview , Amphetadesk’s creator says:

It’s not 1.0 yet because I don’t think it’s great enough for a 1.0 release

I have to agree that there are some yet-to-be-added features (e.g. last-changed-time, summary-with-detail-on-demand vs giant-monolithic-page) that I wouldn’t want to be missing in a 1.0 release, but the quality is great, so I expect it once feature-filled to be solid.

2 comments to “Amphetatesting”

  1. lost some comments in our crash…

  2. There was a comment from Morbus Iff, AmphetaDesk’s creator, answering most of the issues I raised – Morbus went on to hack me a template that does pretty well what I wanted it to. Very nice.