D. J. Bernstein wrote:
> In contrast, sendmail crams all its deliveries into a single host. Do
> you realize what happens when that one host is overloaded?

And you'd rather cram it into a bunch of hosts, which each may be
overloaded and so may the pipe to them.

Patient dead, problem solved.

> > not to mention what happens to some poor country like Kenya.
> 
> What happens to some poor country like Kenya is that qmail uses _less_
> bandwidth, saving money and improving response times. It saves more than
> enough in DNS lookups to wipe out the loss in SMTP compression.

Let's say you have a constrained pipe, and multiple MX for a given
destination.. according to your statement, qmail would open multiple
channels to multiple hosts out there.

And that's efficient?  I don't think so.

It is efficient as long as the pipe and MX range on the other end are
not significantly loaded.  As soon as that changes, your wonderful
theory falls apart.

> Why do you find it so difficult to understand that bandwidth is a
> _shared_ resource? In a typical case, yes, I'm doubling 0.02% of the
> traffic, but I'm also chopping 0.14% of the traffic in half.

What is a typical case?  Your argument is suspended in mid air because
you haven't defined a baseline.

> Latency is different. It is _not_ shared. 

Latency is shared the moment your resources are constrained.  

Again, you seem to exist in an ideal world.  Mind showing all of us how
we all can be enlighted and enter paradise? ;-)

> Every message must get through
> on time.

What is on time?  Again, there is no baseline which allows the
definition of "on time" delivery.

>If a user has a crucial message delayed by two hours, he won't
> care about billions of messages that were delivered in two seconds.

True.  However, that's a matter of mail priorities and resource
constraints.  

Mail priorities largly do not exist or are not commonly implemented in a
compatible way, because there are no real standards out for that.

For an end user, it is extremely difficult to understand and to identify
resource constraints.  Based on that, there are many users who don't
care, nonetheless, it doesn't change anything about the way their mail
is delivered.  qmail won't, and Sendmail won't either.  It's a basic way
of life (by design, I might add) on the Internet.  

Go change/add to the RFCs if you need change and you can round up enough
supporters.

> Fine. Take your delusions to comp.mail.misc; I'll address them there.

If there's someone who is suffering from delusions, it is you, because
a) you don't establish a baseline against which you can measure and
prove your statements, and b) your assume ideal conditions for your
resource considerations.   

Both of it is leathal in the real world(tm).

Products usually win by market superiority expressed in existing market
share {reputation, effectiveness of marketing, etc), financial
constraints.  Bill Gates didn't get anywhere by pissing the rest of the
world off.  He bought them or destroyed them with superior marketing.
However, I suggest you simply try to be better, well founded and
convincing thru logic and not just hype and flaming ppl who have a
different opinion than yourself.  You can gain market share with hype,
but you can only keep it with reputation based on delivering upon
promises.


Regards,
Chris

--
Christian Kuhtz <ckuhtz@paranet.com>
Network/UNIX Specialist
Paranet, Inc.