Re: I-D ACTION:draft-klyne-msghdr-registry-02.txt

In response to no item here in particular, I have .02331 Euro (at today's 
exchange rate) to stick in.
As Larry said, in response to Keith's
> what you seem to be wanting is for IETF to bless nonstandard practices
...we don't want the IETF to do anything but ask the IANA to maintain a 
list of explanations, describing what fields are, in fact, being used, and 
in what ways.
This is much like the argument about whether a dictionary should be 
descriptive or prescriptive -- whether it should tell how words *are* used, 
or how they *should be* used.  Look up "comprise", for instance.  The 
correct usage is "the U.S. comprises 50 states," but most of the time 
you're more likely to hear "the U.S. is comprised of 50 states."  And so 
most dictionaries will tell you that.  But when William Safire writes an 
articles in the NY Times on Sunday, about the word "comprise", he'll tell 
you that the former is right and the latter is wrong.  The dictionary is 
taking the role of IANA, registering "usage", so that when you hear/see it, 
and you look it up, you can see what's going on.  The RFCs are William 
Safire, telling you what you SHOULD do, what you MUST do, and what you MAY 
NOT do.
Let me give a real, practical example of why we need this, from my viewing 
of this very mailing list.  I have a server accumulating these messages and 
presenting them to me through NNTP, and I'm using MS OE to view them. 
Here's a subset of headers from four messages, which I'll refer to in the 
order below as K1, P1, K2, and D1:
> From: Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu>
> Message-Id: 
> <200201312140(_dot_)g0VLecp29397(_at_)astro(_dot_)cs(_dot_)utk(_dot_)edu>
> In-reply-to: Your message of "Thu, 31 Jan 2002 13:07:32 PST."
>    
> <5(_dot_)1(_dot_)0(_dot_)14(_dot_)2(_dot_)20020131130524(_dot_)0324a008(_at_)127(_dot_)0(_dot_)0(_dot_)1>
>
> From: Pete Resnick <presnick(_at_)qualcomm(_dot_)com>
> Message-Id: <a05100301b87f78662990(_at_)[216(_dot_)43(_dot_)25(_dot_)67]>
> In-Reply-To: 
> <200201312140(_dot_)g0VLecp29397(_at_)astro(_dot_)cs(_dot_)utk(_dot_)edu>
> References: 
> <200201312140(_dot_)g0VLecp29397(_at_)astro(_dot_)cs(_dot_)utk(_dot_)edu>
> X-Mailer: Eudora [Macintosh version 5.1fc3]
>
> From: Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu>
> Message-Id: 
> <200201312254(_dot_)g0VMsR900405(_at_)astro(_dot_)cs(_dot_)utk(_dot_)edu>
> In-reply-to: Your message of "Thu, 31 Jan 2002 16:43:18 CST."
>    <a05100301b87f78662990(_at_)[216(_dot_)43(_dot_)25(_dot_)67]>
>
> From: Dave Crocker <dcrocker(_at_)brandenburg(_dot_)com>
> Message-Id: 
> <5(_dot_)1(_dot_)0(_dot_)14(_dot_)2(_dot_)20020131151909(_dot_)032f9e80(_at_)127(_dot_)0(_dot_)0(_dot_)1>
> In-Reply-To: 
> <200201312254(_dot_)g0VMsR900405(_at_)astro(_dot_)cs(_dot_)utk(_dot_)edu>
> References: <Your message of "Thu, 31 Jan 2002 16:43:18 CST."
>    <a05100301b87f78662990(_at_)[216(_dot_)43(_dot_)25(_dot_)67]>
> X-Mailer: QUALCOMM Windows Eudora Version 5.1
These message should have been shown, threaded, this way:
 K1 -> P1 -> K2 -> D1
but OE showed them this way:
 K1 -> P1 -> D1
 K2
(which, by the way, makes it hard to follow the discussion sometimes, 
because Keith's mail, lacking any "References" header field, is always 
shown at the top level of the thread).
What do we see from the above?  Well, that some mailers use X-Mailer, and 
some don't (so I don't know what Keith is using -- presumably it eschews 
"X-Mailer" because it (the header field, not the MUA) begins with an "X"). 
More importantly, we see inconsistent usage of the "References" and 
"In-Reply-To" fields in three different MUAs (two of which are even made by 
the same company, the Windows and Mac versions of Eudora), and we see this 
inconsistent usage's effect on a fourth MUA.
[I'm actually quite puzzled at what the Windows Eudora is doing, why 
message D1 says "in reply to K2", but "references P1"... but that's not a 
subject for this discussion.]
It would Really Be Nice if there were a place implementers could go to see 
how these, and other, fields are meant to be used, both by those MUAs 
generating them, and by those consuming them.
So what am I saying?  I'm saying that I think we should do everything we 
can to (1) document what's out there (not "bless" it, but "document" it), 
and (2) make it easier to document it in the future, and to avoid 
collisions of conflicting usages.  And if that means having 
"X-Stupid-Field" in that list, then that's what it means.  Perhaps there 
can be a clause that allows the IESG to put its comments on any registry 
entry, so that the IESG can label something with "This is a Bad Idea," when 
it finds that appropriate.  It's still better to know than not to.
Barry

--
Barry Leiba, Internet Messaging Technology  
(leiba(_at_)watson(_dot_)ibm(_dot_)com)
http://www.research.ibm.com/people/l/leiba