Re: [Exim] Exim on Cygwin

Top Page
Delete this message
Reply to this message
Author: Phil Pennock
Date:  
To: Pierre A. Humblet
CC: exim-users
Subject: Re: [Exim] Exim on Cygwin
On 2002-03-19 at 21:22 -0500, Pierre A. Humblet wrote:
> Excellent point. Currently I believe there will be a panic die.
> I should fix that.


Hrm. Is there a way to declare some area of a file-system to be
case-sensitive? Or is there another method?

I'm not sure that you can fix this, without moving away from base62
encoding.

The likelihood of a collision, given the process-id encoding, might be
quite low. But given enough mails ...

> By the way, could this happen too if the clock is changed?


In Unix, the "clock" never goes backwards. In theory. The clock is
seconds since Epoch, defined as 1970-01-01T00:00:00Z. Timezones are
handled as a translation-to-human-format thing. Time synchronisation is
normally via something like NTP, or at least some program which calls
the adjtime() kernel interface. That interface redefines the length of
a second :^) by +/- 10%, for long enough to correct an error.

If a sysadmin _does_ set the clock back, on a running system which is in
service, then they're taking a risk. Yes, you could cause an Exim
panic. On the other hand, unless your mail volumes are unusually high,
the expectation of a conflict for any reasonable overlap is quite low.
There are enough other factors in the message-id to minimise the
problem. When you're collapsing half your namespace, the risk is
greater.

If I find someone setting a clock backwards, they get dragged
kitchen-wards for education. :^) If nothing else, people will avoid
doing it again for fear of more education.
--
The pure and simple truth is rarely pure and never simple. -- Oscar Wilde