Here's the HTTP headers I see when retrieving $HUDSON_URL/jnlpJars/slave.jar:

HTTP/1.1 302 Moved Temporarily
Date: Wed, 24 Feb 2010 21:51:10 GMT
Server: Apache/2.2.3
Location: http://hudson/jnlpJars/slave.jar/
Content-Type: application/java-archive

And after following the 302 redirect to $HUDSON_URL/jnlpJars/slave.jar/:

HTTP/1.1 200 OK
Date: Wed, 24 Feb 2010 21:51:02 GMT
Server: Apache/2.2.3
Content-Disposition: attachment; filename=slave.jar
Last-Modified: Sat, 20 Feb 2010 00:28:22 GMT
Expires: Sat, 20 Feb 2010 00:28:22 GMT
Accept-Ranges: bytes
Content-Length: 213820
Content-Type: application/java-archive

I'm writing a custom script to bootstrap slave nodes that essentially calls wget on the master server to get the slave.jar, saves it off somewhere, and runs slave.jar. I noticed the redirect when using wget, because it causes two problems:

1. By default, the file will be saved as index.html instead of slave.jar. This can be overridden by adding -O slave.jar to the wget command-line.
2. Wget's conditional GET support (-N) using the timestamp of the file doesn't work - it always follows the redirect and gets the file, so I'm stuck downloading that 200K slave.jar regardless of whether it's the most recent version or not.

I can work around the former issue and live with the latter, but it'd be nice to have this fixed.