Caleb Clausen (clausen@connsys.com)
Mon, 29 Jan 2001 12:02:58 -0800
one would like to be able to use message-context in a limited user
agent (for instance, one which can only present audio to the user)
to provide the equivalent of multipart/voice-message from vpim v2.
that is, it is an easy-to-evaluate indicator to the receiving user
agent that the message contains data that it can present.
the easy-to-evaluate part is important for performance. the
alternative is to scan through the entire message looking for body
parts with particular content-types.
this type of use seems to be implied by the draft in, for instance,
the following paragraphs:
> The uses of this kind of presentation characteristic for each
> message is multi-fold:
[snip]
> o Filter the message list for presentation via
> limited-capability
> user interfaces (e.g., there is no point in offering images
> when the user is connected by a voice-only telephone user
> interface).
however, there are some statements later on which make this use
impossible:
> o Incorrect or invalid message classification must not result in
> failure to transfer or inability to present a message.
> A receiving user agent MUST NOT depend on the indicated message-
> context value in a way that prevents proper presentation of the
> message. If the value is incorrect or does not match the message
> content, the receiving user agent MUST still be capable of
> displaying the message content at least as meaningfully as it would
> if no Message-Context value were present.
> Message-Context is only an indicator. We do not intend for it to
> convey information that is critical for presentation of the
> message. One can conceive of goofy situations, such as a message
> marked "voice-message" but without an audio body part. In this
> case, the fact that the contents of a message don't match its
> context does not mean the receiving system should generate an error
> report or fail to deliver or process the message.
the draft is saying, in other words, that one cannot depend on the
message-context being correct. if you aren't allowed to depend on
it, why should you even use it at all?
let me give an example to illustrate what i mean. suppose an audio-
only user agent gets a message with a message-context of text-
message but which actually contains audio data. the receiving ua
sees the text message-context, but it cannot use that to filter the
message out of the set of messages it presents to the user since it
"MUST NOT depend on the indicated message-context value in a
way that prevents proper presentation of the message." in other
words, it must still go through the entire message looking at mime
headers for a compatible content-type. if the ua has to look at the
mime headers anyway, then what's the point in consulting the
message-context in the first place?
why can't we just say that the message-context MUST indicate a
certain type of content?
as i see it, either the message-context must be made reliable (if
it's present at all) or we must remove the paragraph about limited
user agents using the message-context to filter messages out.
"For, I say fortunately, I always carry a spare set
of feathers."
--Foghorn Leghorn
This archive was generated by hypermail 2.0b3 on Mon Jan 29 2001 - 22:45:12 IST