On Fri March 11 2005 10:56, Tony Finch wrote:
On Fri, 11 Mar 2005, Bruce Lilly wrote:
In light of the fact that ordering is not significant and
cannot be guaranteed, why does it matter -- how does:
optional-field)
from the same section not work?
Because when generating messages I want to be compliant with the
specification. Robustness principle and all that.
Then when *generating* a message with extension (i.e. not defined
in [2]822) or user-defined (beginning with X-) fields, optional-field
is what applies.
When *parsing* a message, order should not be taken as significant
(because it isn't).
That leaves *modification* of an existing message, which is at best
a gray area. One of the deficiencies of the messaging architecture
as it has evolved is that modification of the message header in
transit is really a layering violation; there really should be a
separate place (call it an "envelope", perhaps) where transport
agents (including list expanders, forwarders, etc.) could place
their markings, separate from the end-to-end, user-to-user message
content (header and body). Something like that could be retrofitted
onto the existing architecture (by wrapping the user-to-user part
in a MIME message/rfc822 wrapper and (MIME) signing that to prevent
modification).