rancid and relays using usercmd patch

As I've blogged before, rancid is a really awesome new cisco config differ.

I monitor about 100 devices, and all was well so far.
Now I've come across a device that I can't reach directly from my rancid host. It's not possible to create a direct network connection, or to use ssh tunnels or netcat to forward those ports using a relay host.
Instead I found a much simpler and more elegant solution. Albeit this involves patching rancid.

TortoiseSVN 1.8.0 and NTLM Authentication on Windows

Subversion 1.8.0 was released recently, and my favorite svn client was updated, too. I had hoped this update to go down smooth, just like going from 1.6 to 1.7 in the past.
This time around there are some authentication issues it seems. TortoiseSVN has dropped neon and relies now solely on serf to handle HTTP(S) authentication.

The release notes mention some server-side configuration changes for optimal performance, but they fail to mention that serf doesn't play nice with NTLM authentication. It does work well with Kerberos though.

So, if you've set up Apache like I described here or here, just add this line to make Kerberos the default and have  TSVN 1.8.0 play nice with single-sign-on.

SSPIPackage Negotiate

This will change the default authentication schema (NTLM) to Kerberos. This will cause TortoiseSVN < 1.8 to prompt for username and password, because the underlying neon library cannot handle Kerberos. TortoiseSVN 1.8 and later will be fine though.
You can force older clients to use serf for specific servers though. I've previously written about it here.

There is a big discussion going on in the subversion-dev mailing list about how to best fix this behavior. It is expected that TortoiseSVN 1.8.1 will include a fix for this issue.

This issue and the fact that most SVN tools still haven't been built on top of subversion 1.8 means that you should stay on TortoiseSVN 1.7.x for now.