I did a checkout for our internal fork of JmDNS [1] and the
upstream version [2] to compare what the differences were..
I found none. There do not appear to be any local edits to
our fork so I can not figure out why we did the fork to
begin with.

Perhaps some older version had locally patched bugs which
have since been resolved? Anyhow, given that JENKINS-8647
is causing lots of people lots of grief, and it is enabled
by default, I went ahead and bumped to jmdns-3.4.0 upstream.

I think reduction of forks for library projects is generally
good, and especially when we have no practical changes made
to our version.

[1] https://github.com/jenkinsci/jmdns
[2] https://jmdns.svn.sourceforge.net/svnroot/jmdns/tags/JmDNS_3_2_1/

(Kohsuke notes: this seeminly no-op change apparently helps Eclipse)

Since we've just changed the group ID and break everyone who depend on
the old artifact GAV, might as well rename artifact IDs as well.
(Otherwise we'd break people twice.)

It looks like this is a left over or a mistake.
We haven't been shipping junit.jar in the hudson.war, and
core has no compile time dependency to junit classes.