Re: linux-ipsec: broken for a few days...

> > > I have just implemented ESP-NULL-MD5-96...

I don't think we need to go even that far. Most real humans using this stuff are going to use Pluto to negotiate their keys and set up their connections, and it suffices to limit Pluto support for it. (People who are doing manual setup can be presumed to know what they're doing.)

(Do note that I say "presumed", not "assumed". They might not. But there are too many ways to screw up a manual setup for us to protect people from themselves very much in that area, at least not if we are trying to build a general-purpose tool.)

Note that we can't claim IPSEC conformance if we make ESP-NULL too hard to get to, since it is mandatory, like support for AH-only connections. But I do think John's initial suggestion is good: it should be turned off in the default configuration. (And possibly our docs should warn against turning it on unless you know what you're doing.)

Hmm, given that AH authentication is really not much of an improvement on ESP's built-in authentication, and an AH-only connection is just as insecure as ESP-NULL, should AH likewise be turned off by default, and warned against? It seems to me that one follows from the other.

                                                          Henry Spencer
                                                       henry@spsystems.net
                                                     (henry@zoo.toronto.edu)