On Thu, 11 Mar 2004, Fred Viles wrote:
>
> Philip's proposal to log both start and end times does address the
> problem that a single timestamp must misrepresent queue residence
> time on one end or the other when transmission time is long.
Note that Exim already does this, albeit obfuscatedly. The message-ID
includes a timestamp in base62 format, and it is created at the start of
message reception. The <= line is logged when message reception has
completed, so the combination of its timestamp and message-ID can be used
to work out the transmission delay.
--
Tony Finch <dot@???>
http://dotat.at/
