Amedeo Rinaldo:
> Il 26/01/2012 13:15, Wietse Venema ha scritto:
> > Do you have quantitative evidence, based on actual
> > email traffic, that this would make a difference?
>
>
> Hi Wietse, thanks for your reply ..i've taken some time to collect some
> data.
>
> I've monitored for 2 days one single host wich handles about 200 domains
> (it cleans mails for ..and forward to external mbox server). I've
> noticed about 2000 remote mailbox verify tasks; redundancy checks
> "caused" by case-sensitive re-checks are about 65-70%
> About every mailbox has 2-3 different case variants.
Are you talking about sender or recipient address verification?
I suspect it is the latter.
According to sender verification logs for the last months on my
small server, case folding the domain part would reduce the number
of unique senders from 7436 to 7433 (0.04%), and case folding the
entire address would reduce the number of unique senders from 7436
to 7419 (0.23%). Most of these sender addresses are spam, and they
are used only a few times.
The situation with recipients may be different, because the same
recipients are used many times. My population of recipients is too
small to make definitive statements. According to this sample, some
senders upcase the first character of the recipient's localpart,
all or part of the recipient domain, or the entire recipient address.
Case folding the domain should be safe because it is defined in the
RFCs as case insensitive, but the safest approach is not to fold
case at all unless it is known to be safe.
This means that case folding can't be on by default in Postfix, and
if it is made configurable, there should be separate controls for
localpart and domain, and for sender and recipient addresses.
Wietse
