duplicate keywords: to be or not to be?

[Dave Kormann <davek@research.att.com>]

 > Supposing the #$#say is considered valid, but that the message is
 > really a single line message with a heap of junk following.  Then you
 > don't set up a bucket to pour all the incoming '#$#& abc' lines into.
 > The * and END handlers then just drop any data they recieve on the
 > floor when there's no bucket waiting to be filled.  It's easy to throw
 > stuff away in this case.
 > 
 > Or suppose it's invalid. Then you don't set up a bucket to ...

yep, this is true; it wouldn't require any real changes to a compliant
implementation to Do The Right Thing; however, i'm becoming convinced
this isn't a terribly useful addition to the spec.  are we aware of
implementations which do send duplicate keywords?  can their authors
be beaten into submission?  it's beginning to seem best to simply
label behavior like this as a bug and punt it, while encouraging
implementors to provide debugging modes which make it easy to spot
problems like this.

 > 
 > I do not see what advantage is gained by supporting duplicate keywords,
 > short of humoring clueless 3rd part implementors.  Um, like me I
 > guess...

so, given erik's note that he vaguely felt these should be dropped
(though he said he didn't have strong feelings), that's two votes for
just dropping them.  votes for keeping them?  any strong feelings?

again, however, if there's a real feeling this should be in the spec,
i'm cool with it.  the text is already there.
				dk