Hmmmm... Adding one of those 1 pixel transparent gif to all my posts, should let me track where my posts are syndicated with a simple logfile analysis...

It's true, Dave has been trying to convince me to write on my own weblog for a while. I particulary remember him telling me to do so while having a good dinner in Trieste, in Marc Canter's favourite restaurant there (this IS a small world!). To be honest I thought that it required too much time and would not bring much benefits. I was wrong: it does require some time, but there are a lot of benefits.

Now, about thinking in the open: I think that we developers of tools and applications based on UserLand's software should start talking a little bit more between ourselves. Playing a little bit more with the idea of Radio or Frontier as an OS, I'm sure that there are lots of opportunities in this community and probably we can help the whole process. How can UserLand help us more with our development or our marketing? How can we help UserLand to sell more and make even more cool software? Let's just not make UserLand another Microsoft or Apple

InstantOutliners are a great working tool (and SharedOutliners will be even better ;-), but the process needed to check updates of subscribed outlines can be pretty bandwidth intensive. André has probably found an at least partial solution to this problem, and I had a confirmation of this issue again this morning.

To check if an outline has been updated, Radio reads the http headers of the opml file you have subscribed to, checking the size of the file, if the file has changed in size Radio interprets this as a change and downloads the opml file.

Now, this polling process is done using a GET on the file, instead of using the HEAD method. Also if Radio is not reading the contents of a file if the http headers have not changed, it looks like using the GET method triggers the file transfer anyway.

Here's how my network statistics app was showing this morning:

See the spikes? No other app was running except for Radio, and they match the polling process.

After changing the polling code to use the HEAD method, this is what the same utility was showing:

This is totally empirical, but also quite evident. I'm not suggesting how to change your polling script here because most probably UserLand will update it soon and because you could break it :-).