Project Status
Date: Mon, 6 Nov 1995 10:53:05 -0800 (PST)
From: Mark Crispin
Subject: Re: Mailer leveling
Ken -
Thank you for reminding me. I have now looked into what needs to be done.
There is no way to do this in any kind of clean way. The code that checks for
login validity has no way of sending messages to a client, it can only return
"yes" or "no". This is by deliberate design; you don't want to give crackers
any information.
The login validation modules are modified for UCS systems only so that if the
"#" character appears anywhere in the shell name, it causes a failure. The
problem is, that way of doing it is isn't extensible to giving any kind of
message. Since the login validation code also supports POP, it can't just
have IMAP wired into it.
So what we're talking about is removing the check in the login validation
sources; and adding an extra getpwnam() call in the imapd, ipop2d, and ipop3d
main program sources to do the check and messages.
Before I go into this kind of mess, I'd really like to ask you to reconsider
if you really want to expire passwords and/or disable access at all for email
only access. In effect, in this state you accept mail but withhold it from
its owner.
I have ethical qualms; that is alleviated since it isn't *my* machine, but I'm
also worried that it may be illegal (I'd certainly talk to a lawyer if it ever
happened to me!). Am I indemnified if the rubber chicken hits the fan? I
know that normally state employees are, but does that cover acts with prior
knowledge of them being questionable? I'd like to ask management to clarify,
and take responsibility.
A possible solution would be if sendmail were modified to bounce mail to an
expired/disabled user. This may be doable in tmail.
By the way, I now know why there were corrupted mail files on mailer*.u the
other day. It turns out that Eliot and I had both forgotten that mailer*
needs to have a different version of imapd than other AIX machines. In
particular, what got installed there was an imapd that has fcntl() locking
disabled.
So it's another reason why I have a problem with this...
-- Mark --
Date: Sun, 5 Nov 95 15:21:00 -0800
From: Ken Lowe
Subject: Mailer leveling
Okay, the mailers should now be even based on the amount of connect time
users are using.
-Ken
-----
Connect time in hours
mailer < 1.0 < 2.0 < 5.0 < 10.0 >= 10.0
------ ---------- ---------- ---------- ---------- ----------
1 2344 0.17 96 1.40 75 3.17 44 6.46 8 14.86
2 2348 0.16 96 1.40 75 3.26 44 6.67 8 21.89
3 2348 0.15 96 1.40 75 3.21 44 6.56 8 16.39
4 2348 0.15 96 1.34 75 3.27 44 6.70 8 13.77
5 2348 0.11 88 1.44 68 3.13 43 6.51 5 15.33
6 2348 0.12 94 1.39 71 3.09 42 6.83 4 21.95
7 2348 0.12 93 1.40 73 3.06 42 6.41 3 11.27
8 2350 0.12 94 1.41 73 3.32 42 6.64 6 15.53
9 2350 0.10 93 1.37 74 3.24 42 6.65 7 13.85
10 2350 0.10 94 1.42 73 3.14 44 6.59 6 17.89
11 2348 0.10 94 1.39 75 3.12 42 6.74 6 13.33
12 2348 0.10 94 1.38 73 3.37 42 6.49 8 14.00
13 2350 0.15 94 1.38 73 3.06 42 6.44 6 14.10
14 2350 0.15 94 1.45 72 3.31 42 6.60 6 12.19
15 2350 0.12 94 1.40 73 3.00 42 6.52 6 12.59
16 2350 0.12 94 1.35 72 3.24 42 6.76 6 12.39
------ ---------- ---------- ---------- ---------- ----------
TOT 37578 0.13 1504 1.39 1170 3.19 683 6.60 101 15.16
AVE 2348 0.13 94 1.39 73 3.19 42 6.71 6 15.94
Back to AST Home Page