Nathaniel -- Your motivation is fine, and is still intact.
The problem is that Message-ID has a rather restricted definition and
narrow semantic in RFC-822, and in normal language. Yes, it is indeed
a globally guaranteed unique header-field label, which is the right
basic semantic, but is also a big part of the problem.
Our "bodypart label" only requires local unicity (within a given message).
Given the length and program generated style of the currently used
Message-ID, I can see difficulty with using it in part A to reference
part B, which is what you need to do to describe aspects of parallel
behavior.
Otherwise you are entirely correct, in that technically it will work.
I think the problem is much more rooted in how we feel about the names.
At this point, we are beginning to define a user interface. Human
users are going to see and use, and manipulate these attributes and
values when they create messages (if they don't when they display and
read messages).
It is this User Interface aspect that is giving us trouble.
We would rather see a new attribute (header) with a new and more
naturally meaningful name for this new "labeling" purpose, than to
conserve on namespace consumption by reusing a name that does not seem
to mean what we want it to mean.
Which leads me to a real question. Is there a Real Estate shortage in
the attribute name space? Do we have a budget limit on the number of
attribute (header) names we can (should) define?
Why should we mentally turn ourselves (and millions of potential
users) inside out with reused names that don't "mean what they mean"?
Life in this new world is crazy enough as it is, without playing mind
games with naming things. We cannot even keep them straight in our
own limited community of experts!
Think in terms of some User Trainer trying to explain this to a room
full of New Users from the Personnel Department.
Better yet, think of Yourself trying to explain this to a room full of
New Users from the Personnel Department.
Now do you get the right sensations? YEAH! I hope so!
NOW, think about writing the users manual for your new UA!
The simple technical fact that our whole basic message structure is
recursive, and is thus technically very powerful does not help to
teach people how to use Message-ID for Bodypart-Label, because most
users are not ready to understand all this recursive WKBCS. Most
users will never be ready for all this WKBCS.
Cheers...\Stef
PS: WKBCS = (Well Known Brown Colored Substance).