Re: [exim] Re: Bug#270735: exim4: Self-denial of mail servic…

Pàgina inicial
Delete this message
Reply to this message
Autor: Greg Kochanski
Data:  
A: V. T. Mueller, Continum
CC: exim-users, 270735-forwarded, Greg Kochanski, Andreas Metzler
Assumpte: Re: [exim] Re: Bug#270735: exim4: Self-denial of mail service
V. T. Mueller, Continum wrote:
> Hello,
>
>
>>On 2004-09-09 Greg Kochanski <gpk@???> wrote:
>>
>>>Several of the settings in exim4 can easily lead to accidental
>>>denial-of-service attacks on one's self.
>
>
> Hm. Could it be that you also disconnected your PC's power switch
> from the mainboard since it could cause computation to stop? ;-)



True, but not all that funny.

>
>
>>>Specifically, setting deliver_queue_load_max, queue_only_load,
>>>smtp_load_reserve all will completely stop mail delivery if a user
>>>starts up some CPU-intensive program.     And, this can happen
>>>easily -- in a scientific/university environment, it's not unusual
>>>to start up a calculation that will take CPU-days to complete.

>
> [..]
>
> These services normally run on dedicated computer clusters.


In your dreams, or mine.
Some of us do computations on relatively normal Linux PCs.
If you go to http://kochanski.org/gpk you can see a publication list
with a reasonable number of papers on it. None using any formal
cluster.

The trouble with "official" clusters is that they get loaded up.
You may be 12th in a queue to get the cluster, so overall,
you end up getting far less CPU time than you might imagine.
*And* you need to do paperwork or politics to gain access to the cluster.
Certain problems need big clusters, but lots of computation
can be done on a single PC, given a bit of patience.



> ... It would
> appear odd to me if these machines would also be used for handling
> of SMTP messages. Basically, it doesn´t matter if the machine handles
> only locally generated messages or does other mail-related jobs, too.
> If email is of importance, then it´ll be the admins job to take care
> for proper settings in the MTA configuration.


Well, my home machine is used for both.

The point is that software should be designed so that
misconfigurations cause the minimum damage.
A subsidiary point is that you are making strong assumptions
about who uses your software and how they use it,
and those assumptions may not always be true.


>
> Maybe there are other opinions out there, but I personally welcome
> the unix-style behaviour that lack of thinking outside the computer
> is not being compensated on the inner side. Probably we will have to
> let go the idea of systems that do strictly what they´re "told" to
> when compute power and software engineering have evolved accordingly.
> Until that day I appreciate every system that bites my butt for every
> single error I make. At least every system that also allows me to
> clearly track down what caused the mishap.


What would life be like if everything ran that way?
Frightening!

I agree that error reporting is crucial, and the ability to
find out what is wrong is crucial, but I can do without the
bites in the butt. I need to sit on it, after all.