bittorrent and rss. Since first posting about it, I've been thinking more and more lately about the possibilities of combining rss enclosures and bittorrent ad-hoc p2p networking. The more I analyze the setup the more I like it.
[Adam Curry: Adam Curry's Weblog]

Oh, it's a great idea.  I'm all for it.  It will do for TV what the Web did for publishing: lower the barrier to entry to enable everybody to do it.  (A $200 Walmart computer, a $40/mo cable modem and a $50/yr subscription to Radio Userland, and you're a publisher.  Add some Digital Video gear to the mix and now you're a TV studio.  Totally awesome.)

I'm definitely not trying to put out stop energy here, but there is at least one problem, and it could be a show-stopper.  In some circumstances, BitTorrent can totally slam your Internet connection, making it unusable for other (interactive) tasks.  This appears to be unrelated to the amount of bandwidth BitTorrent is using.  BitTorrent's built-in throttling -- which is ingenious -- doesn't seem to help here. 

I suspect it has something to do with one's uplink hardware (cable modem, etc.) getting slammed by extraneous incoming requests.  These extra incoming requests are not serviced, so it's overhead traffic.  I don't have any proof that's actually what's going on, but that fits my observations.  I'll develop huge latencies (frequently over 1500ms) that go back to normal (20-40ms) as soon as I kill BitTorrent processes.

The point: the system Adam describes would work in the background without user intervention.  Due to the side-effects outlined above, it might be wise to add controls which would allow the user to disable BitTorrent in the event it causes network performance problems.

4:50:42 PM   

Even Sun Can't Use Java "It turns out that Sun does not eat its own dog food. Specifically, this internal memo from Sun strongly suggests that Java should not be used for Sun's internal projects." [Slashdot.org]

Oh, man, for years I've been complaining about Java's virtual memory usage, saying that the Java VM's memory allocation behavior is pathologically bad.  See the memo's point #2, "The JRE is very large."

This memo should not be seen as an indictment of Sun.  This sort of internal dialog is (in my opinion) healthy.  It's good they're articulating the problems they're running into as they gain experience deploying and maintaining Java in production systems.

As a developer who has been deploying and supporting Java-based solutions for years now, I can confirm that all of the complaints outlined in the memo are valid in general, and apply specifically to non-Solaris JREs as well.  (Where the memo says the problems are worse on Solaris, I'll have to take their word for it.)  I've felt their pain -- this ain't baseless political bickering.  ISVs deal with these problems all the time.

I'd also like to point out that many of these problems may not be exclusive to Java.  As the .NET platform matures it may also suffer these limitations.  Only time will tell whether Microsoft will do better.  In other words, I don't think any of these complaints can be used (today) as reasons to switch to .NET from Java.  Others will disagree.

1:46:50 PM