Changing passwords - rationalisation of methods?

[Darren Reed (Optimation)]

As part of the stated direction to move the MIT Kerberos
effort towards being "clientless" (or at least without the BSD
client applications), is there some consideration being given
to rationalising the applications for changing a password?

What I'm driving at here is that there are two kpasswd's and
a v5passwd & v5passwdd pair.  The two kpasswd's are not
nearly identical in code despite their near identical man pages
(or maybe it is the subtle differences that are important).  At
least v5passwd seems to have a specific purpose.

Are all three really needed?  (kadmin/passwd would be the
I'd favour for dumping.)

Darren