From: Brad Templeton (brad@templetons.com)
Date: Mon Jul 02 2001 - 04:18:34 CDT
On Mon, Jul 02, 2001 at 09:26:21AM +0100, Clive D.W. Feather wrote:
> Brad Templeton said:
> > Actually, I can't recall the last time I saw a message/partial in
> > mail, but they are very common in USENET. Message/partial is far
> > more USENET's issue than Mail's.
>
> I could say "nonsense", but I'll merely say that my experience is
> completely the opposite. I don't read any groups where message/partial
> occurs, whereas I regularly see it in mail. (Turnpike splits large messages
> into pieces and reassembles them (correctly).)
Well, I'll retract a bit and say my brain was thinking "multi-article
message" when I wrote message/partial above.
And it remains true that I rarely see multi-part e-mails, in fact perhaps
I'm just on the wrong mailing lists but I have never received one to my
memory, and there are of course tons of multi-part messages on USENET,
mostly uuencoded.
If message/partial is that rare in news, then there are no existing practice
issues to deal with regarding it. We can freely right what the right way
is, under our spec, to post messages of various size, and set a
threshold where the poster should expect all tools to be able to handle
a complete message, a range of doubt, and even a range where at this
time message/partial is likely necessary, though advising that number will
move upwards.
I still maintain that large message assembly and disassembly is a transport
layer issue, and should only be dealt with with great reluctance in the
article format.
And yes, it can be true on some lines that two sessions are more efficient
than one, but frankly, that needs to get pretty bad before applications
should start doing transport things to make up for poor transport
implementations.