suggestion for locating master kdc logic

[Nico Williams]

Will, I think you could argue that the change required in Solaris
Kerberos to match MIT krb5 behavior in this case is a bug fix, not a
backwards-incompatible interface change requiring a minor release
vehicle.  If the ON c-team and/or PSARC disagree then you'll have to
stick to the divergence in behavior -- I don't think you should expect
MIT to "fix" this, because I don't think MIT should.

This may be a classic case of release schedule mismatches between
upstream and downstream.  This is something the upstream should
generally avoid doing to known, friendly downstream consumers, but
it's difficult to prevent entirely, particularly when the upstream is
less of a stickler for
behavior-that-rises-to-the-level-of-an-interface than PSARC might be.

Nico
--