Re: Negotiated noncompliance

[Barry Leiba <[e-mail]>]

Dave writes:
> At 01:32 AM 8/17/00 -0700, Eliot Lear wrote:
> Implementors do not need language in a standard to tell them that they 
can
> do things differently if they know what they are doing.

I usually find myself in agreement with Dave, but not this time: I agree 
with Eliot that because of the situation that exists now, in deployed 
servers, it's a good idea for us to say *something* about it.  Absent that, 
what happens (you know it; you've seen it) is that implementors see 
noncompliant stuff coming in and they simply change their implementations 
to accept it -- often without realizing that this might have undesirable 
effects.  Wouldn't it be a good idea for us to give them some guidance, 
based on, as Eliot says, "the experience of nearly twenty years of odd 
interoperability failures".  I think so.  We can massage the wording, but I 
think it's not the best answer to say what Dave says above.  The fact is 
that many, many implementors *do*, in fact, need wording to help them out 
here, because they often *don't* know what they're doing -- not to the 
extent that those of us with many years of experience do.

I do agree with Dave in this:
> Absent the details of what constitutes "great care", what does a reader
> learn that they did not already know?

Right... If we're going to say something like this, it's probably a good 
idea to say why -- perhaps in a separate document (though I'm not sure how 
DRUMS would do that).

Barry Leiba, Internet Messaging Technology   ([e-mail])
http://www.research.ibm.com/people/l/leiba