On Thu, May 24, 2012 at 12:04:30PM -0400, Greg Hudson via RT wrote:
> We have someone interested in solving this issue.  Do you think it would 
> be sufficient to add an inetd option to krb5kdc?
> 
> Because we serve over UDP and TCP, and because correctly implementing a 
> UDP server in the Unix socket interface is more difficult than it should 
> be, our server network loop is very complicated.  Deferring this issue to 
> inetd would have a small UI footprint and would add a small amount of 
> additional complexity.  Adding specific address configuration would have a 
> larger UI footprint and a larger increase in complexity, I expect.  But 
> it's something we could still consider.

It sounds like it could.  Would such a setup end up firing up a
different KDC (or kadmind) process for each listening address that
received traffic?

Nalin