- From: Keith Moore <moore@cs.utk.edu>
- Date: Wed, 22 Aug 2001 17:20:50 -0400
- To: Patrik Fältström <paf@cisco.com>
- cc: Keith Moore <moore@cs.utk.edu>, Eric Brunner-Williams in Portland Maine <brunner@nic-naa.net>, jpalme@dsv.su.se, discuss@apps.ietf.org
> I was not talking about NAT's, but things that block traffic on certain > ports, like normal firewalls, but you are completely right that this can be > used for NAT purposes aswell. > > But, I get your point. Doing DHCP request, pppoe authentication etc when a > host "wakes up" and get's an IP address is one thing. Doing the same or > similar things when it for example starts it's "SIP telephony listener" or > initiates some other flow is not good. > > That is what I read in your message. That's true, but there's more to it than that. Yes, doing a dynamic DNS update on a per-flow (rather than just a per-login) basis is a pain. But more generally, expecting all apps to use SRV on a per-connection basis is a bad idea. DNS names are ill-suited as endpoint identifiers for a variety of reasons, including that DNS lookups are slow (often taking several seconds to complete), they're not terribly reliable (servers often falsely report errors and/or are misconfigured) Putting SRV lookups in the path slows things down and makes the overall app less reliable. Port-hopping also makes it more difficult to diagnose and fix problems. You might get this to work for specific applications, but it's not a general purpose solution. And the last thing we need is for the network to become more application-aware and application-specific. Keith
Received on Wednesday, 22 August 2001 17:22:57 UTC