On Mon, 12 Mar 2001 18:41:15 PST, "James P. Salsman" said:
Consider multiple devices producing or accepting the same media type. For
example, if I send you an email with an audio/basic attachment, is there
value in your knowing whether it came directly from my microphone or off
of my file system? I think there is. I would really like to know what
you think. Is there some better way to communicate the source information?
Nope. No value in knowing. At least not at the level that a single parameter
gives you.
First off, "directly from my microphone" is *highly* unlikely to be
incredibly factual. More likely, as you compose the message, you decide
you need an audio clip, and when you record, it goes into a temporary
file on disk before being base-64'ed and included as part of the outbound
stream. It's *conceivable* that your MUA reads directly from your
system's equivalent of /dev/microphone, base64's and writes directly to
the SMTP datastream outbound, but that is (a) painful to contemplate
and (b) makes the attachment of multiple bodyparts "interesting".
So we'll take it as a given that the audio *did* (at least potentially,
for systems that have in-memory file sytstem caches) hit your filesystem
at some point.
So... what you *really* care about, I suspect, is "was this audio
clip recorded while composing the message" or "is this a clip from
yesterday he attached". What you want is a *datestamp* of when
the recoring was made, and an identification of *who/what* it
is a recording of (is it me saying it yesterday, or a sound bite
of something Steve Bellovin said at an IETF plenary a while ago)?
In addition, even if it was recorded "right now", you need to have
identification if the file was edited (even if it was merely to
remove dead air, pauses, stutters/stammers, background noise, etc).
You also need to identify if the audio is a composite mixed from
several sources (I could create a clip of me speaking with a
background of a MIDI transcription of a Bach concerto - was this
from the microphone that my voice came from, the software synthesiser
that created the Bach, or the software mixer that combined the two?
What if it's a composite of several different audio clips? I've seen
rap songs that the credits run to several column-inches due to the
heavy use of sampling.
And of course, there's the *biggest* issue - if it's *really* important
to know whether it came off the microphone or file system, how do
you *verify* the field values? Lest you think this is purely academic,
I suggest poking around the Bugtraq archives and seeing how many CGI packages
have broken because they blindly trust 'HIDDEN' form input "because nobody
would change the price of a CD player". Or Napster offering to filter
filenames "because nobody would ever misspell Metallica".
Other than that, I don't see any problems. ;)
Valdis Kletnieks
Operating Systems Analyst
Virginia Tech