On Tue, 13 Jul 2004, Martin Stecher wrote:
In earlier discussions we found that MIME is a good fallback when
negotiating profiles and no match for a more detailed application is
found. So, exploring a MIME profile indeed has some value.
On the other hand I think that we must concentrate on SMTP if we
want to do OPES for email not MIME or any other email protocol.
Most messages in the Internet are transferred using SMTP. POP
and IMAP are used to fetch them from a mailbox but they reach
the mailbox usually via SMTP.
Is it reasonable to assume that, due to differences in user agents
(e.g., PC versus phone), folks will want to adapt stored messages when
they know IMAP or POP client preferences (rather then when the message
is received by SMTP server)?
And the SMTP dialog has information that is needed as meta data in
addition to the MIME message for effective filtering, for example if
you want to check the recipients of a message you better ignore the
MIME header but look at the RCPT TO parameters from the SMTP dialog.
Yes, it is clear that MIME alone is not sufficient for some (most?)
adaptations of e-mails. It is only a common denominator among SMTP,
ICAP, POP.
In addition we should consider that some OPES services may want
to adapt the SMTP dialogs, for example control the relay behavior
of an MTA (decide which sender is allowed to send to which recipients).
That is definetly beyond MIME message filtering.
This is very important, I guess. Should we decide now whether SMTP
commands adaptation is in scope (as opposed to SMTP message
adaptation).
So, I think we should concentrate on OCP/SMTP. As adapting MIME
messages is the most important sub part of that we may want to
explore that first and design the profile in a way that allows to
create an OCP/MIME profile easily by simply deleting all parts that
are beyond the MIME message part.
I see your point (and glad that you are back from vacation!). I have
not decided yet. What are we really trying to optimize here: Do you
think we will waste a lot of time if we define a common OCP/MIME
profile (that alone will not be useful for some adaptations) and then
build OCP/SMTP profile on top of that (as opposed to building one
monolithic OCP/SMTP profile)? Or do you think that defining a common
OCP/MIME profile is simply useless because nobody cares enough about
IMAP and POP (and so nobody will reuse the common profile).
Thanks,
Alex.