[2621] in SIPB_Linux_Development
Re: MIT Mailhubs Linux-Athena mail from outside net-18
daemon@ATHENA.MIT.EDU (Derek Atkins)
Sun Feb 14 13:43:05 1999
To: Karl Ramm <kcr@MIT.EDU>
Cc: linux-dev@MIT.EDU
From: Derek Atkins <warlord@MIT.EDU>
Date: 14 Feb 1999 13:42:22 -0500
In-Reply-To: Karl Ramm's message of 14 Feb 1999 11:34:20 -0500
I claim that it is a bug because it is a radical change in (long-term)
behavior with little to no warning. I personally have successfully
been using the MIT mailhubs to send mail "from MIT" for a number of
years, and only in the last week+ have I seen this problem. Moreover,
I am not the only person to notice the problem.
Unfortunately there is no way to know whether the installed machine is
'on-campus' or 'off-campus' when it is being installed. This is the
same discussion we had on zephyr yesterday regarding /etc/resolv.conf
-- there is no way to know during the install, when the particular
package is being loaded onto the system, where the machine is going to
sit on the net. Also, if someone moves off-campus, their mail will
start bouncing.
Yes, I realize that SPAM is an issue. I understand the concept of
'open relays'. However, I also believe that 'open' means sending mail
from anywhere to anywhere through that host. I'm not asking for that.
I'm asking for mail to be sent 'from MIT' (although connecting from
anywhere) to anywhere. This is different than an open relay.
-derek
Karl Ramm <kcr@MIT.EDU> writes:
> (postmaster nuked from cc: line)
>
> So, you seem to be saying that you tried to send mail from off campus to
> off campus through the campus mailhubs and it didn't work, and that this is
> a bug @i(in the mailhubs). This turns out not to be the case in the modern
> internet of spam sent through open relays.
>
> I don't think we really want to have mail queuing (for inaccessable hosts)
> on users machines in the dorms, which is what would happen if we implemented
> your proposed solution in the general case.
>
> Maybe an option in the install for off-MITnet machines that put in a more
> suitable sendmail.cf? I don't know the details of the redhat install
> infrastructure, but maybe you could tie it in with an option that brings as
> much local as possible...
>
> kcr
--
Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
Member, MIT Student Information Processing Board (SIPB)
URL: http://web.mit.edu/warlord/ PP-ASEL N1NWH
warlord@MIT.EDU PGP key available
