Apparently I haven't hated on GNOME recently. It's long overdue.
So Gossip is currently my preferred Jabber client. It's simple, gets the small chat window thing down, and generally stays out of my way (no, I don't want to be interrupted while I'm working).
Most of my hate for Gossip comes in how it handles logs.
First, it logs in a standalone XML format. Personally I'd prefer plain text, but I may be in the minority here. Whatever, maybe this will piss me off enough to learn XSLT.
Second, when you request to view a chat log Gossip loads the entire log in a Web browser. Some of my logs are at least a few megabytes in size, so loading them takes 10-20 seconds.
Third, Gossip uses XSLT to convert from XML to HTML. This process leaves a temporary file around. Closing Gossip doesn't seem to remove any of these files. Last time I checked I had around 100 gossip-log-XXXXXX.html files polluting /tmp.
Finally, Gossip uses gnome_url_show to open the log files. Ostensibly this is because GNOME has some knowledge of your "preferred" applications, but it doesn't work.
Neither of my GNOME installations have a handler for file:// URLs. So on my home computer Gossip opens the logs in Epiphany, while at work Gossip opens the logs in Firefox. Hey, shithead, I use Galeon.
So my "solution" to this is to add GConf keys for file:// URLs:
gconftool-2 --set /desktop/gnome/url-handlers/file/command --type string "galeon --noraise --new-tab %s"
gconftool-2 --set /desktop/gnome/url-handlers/file/enabled --type boolean true
I dunno, this may invalidate my GConf schema. There may be a way to do this using Nautilus or some other hateful software, but this works for now.
Either I don't understand RSS 2.0 TTL, or Bloglines is behaving worse than I thought. My feeds have the TTL set to 60 minutes.
Bloglines is actually fetching the feeds every half hour.
I'll be bumping that to 180 or 240 tomorrow, at least for fucking Bloglines…
So most of you probably know about Podbop by now. I had a really good time in California, especially in San Diego visiting Eventful. Taylor's summary has pictures, so yeah.
I'm still amazed at the the response we've received so far. Many passionate users have emailed us with suggestions (special thanks to Tim!). I'm working on some of the ideas as I have time, but the list is long and I have a full-time job. ;-)
Unfortunately I can't give you stats on how many people have subscribed to Podbop feeds. Looking through the server logs, however, I'm a little concerned at the behavior of one user agent: Bloglines.
You see, Bloglines likes to crawl the 40 or so Podbop feeds its users have subscribed to as quickly as possible. Most of the time, it issues all 40 requests in under ten seconds. To me, this seems irresponsible.
My little Linode doesn't have the processing power to serve Bloglines' average of four requests per second on top of the usual load. So should I just serve 503 Service Unavailable to every other Bloglines request? Or can I suggest that Bloglines spread out its requests in some way? I guess I could adjust the feed TTL, but is there a better way?