mail-ng
[Top] [All Lists]

Re: Delete all and replace is practical

2004-02-01 09:01:22


On Jan 31, 2004, at 11:20 PM, Dave Crocker wrote:

iceburg. Personally I think that having all those IP routers doing store and
forward is a far greater source of problems.

Let's just fully interconnect every host.

Never going to happen. Any protocol that demands it is dead in the water. corporate security issues will kill it.

<Sermon>


Let's not complain about intermediation points.  They have provided
interesting careers for most of us.

</Sermon>


Amen. Especially since they aren't going away.

<Sermon>

I also think there's some inherent assumption among some folks here that this new thing is going to be Unix/Linux based. We need to remember that not only isn't the existing reality all Unix/Linux, we're designing for the future, where we have no idea what the underlying infrastructure will be, either on the network or hosts.

That's a very important reason why we need to focus on protocols, services, api's and interfaces and not what goes on behind that, because this thing needs to survive beyond the internet, beyond TCP/IP, and beyond our current OS systems. It's also a reason why we should avoid growing the charter into the storage and delivery parts, because it's not just a case of grabbing Cyrus iMap and modifying it to the new setup... You don't know what the filestore and its interface are going to be, or even if you have a filestore.

SMTP focusses on the network aspect of this, port to port, and necessary services to support that. And we should keep that same focus here, and leave what happens after a message is accepted to some other thing. It's a big enough challenge just to fix that piece that taking on new challenges will guarantee failure, or significant delays, which given the current state of SMTP, are as bad as failing.

</Sermon>

Now be careful out there. Here there be Tygers.