RE: [exim] smtp_accept_max_per_connection

Top Page
Delete this message
Reply to this message
Author: Timothy Spear
Date:  
To: exim-users, 'Bill Hacker'
CC: 
Subject: RE: [exim] smtp_accept_max_per_connection
I agree with Philip, you are adding significant project cost with minimal
gain for the system. When you consider the cost in terms of complexity,
maintenance, new bugs, additional memory space for the code to handle the
additional complexity.... it will very quickly add up. Lastly, looking at my
configuration, I do not see an extensive you of Boolean variables which have
a life beyond the immediate equation evaluation or at most the processing of
a single message. We would need the volume of AOL or Yahoo to make the
memory and processing savings worth it.


Tim



-----Original Message-----
From: exim-users-bounces@??? [mailto:exim-users-bounces@exim.org] On
Behalf Of Philip Hazel
Sent: Wednesday, November 30, 2005 10:09 AM
To: Bill Hacker
Cc: exim-users@???
Subject: Re: [exim] smtp_accept_max_per_connection

On Wed, 30 Nov 2005, Bill Hacker wrote:

> Given that there can be several hundred simultaneous connections,
> - each allocated ten acl_c(n) variables,
>
> and that each connection could process multiple messages,
> - each allocated ten acl_(n) variables,


... but only one message at a time, so this does not make any
difference, at least not for reception. It is true that the delivery
process for message number n may still be running while message n+1 is
being received, but that is the same whether the two messages came on
the same connection or on different ones.

> and that the curent variables are not simple booleans, but capable of

holding
> potentially lengthy strings.......


No difference (on a 32-bit system) unless the variables are actually
used. An unused variable occupies just one word - a NULL pointer.

> Might it not be more economical to add a set of variables of a different

type?

I really don't want to get into typed variables in Exim. It's a whole
new ballpark. And I don't think current systems are so short of memory
that we have to do drastic things like store Booleans as single bits.

Doubling the number of variables would add 80 bytes to the memory used
by an Exim process on a 32-bit system, if those variables were never
used. This is like adding one additional header line to every message,
which is (IMHO) not a lot.

-- 
Philip Hazel            University of Cambridge Computing Service,
ph10@???      Cambridge, England. Phone: +44 1223 334714.
Get the Exim 4 book:    http://www.uit.co.uk/exim-book


--
## List details at http://www.exim.org/mailman/listinfo/exim-users
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://www.exim.org/eximwiki/