On Jul 3, 2021, at 7:47 AM, Eliot Lear <l...@lear.ch> wrote: > I don't think Tim could be blamed for holding the view that there is a > separation between specifications and how they are used. There's good and bad > to the practice. The good is that the spec can be used in ways that the > creators didn't intend, and thus perahsp there are fewer unnecessary > constraints. > > On the other hand, not having a theory of operation section, as we do have in > a good number of our specs, leads to people really not understanding when > they are applicable, and perhaps more importantly, when they are not.
People don't even understand how to use the specs as intended. We're essentially telling people "EAP methods are applicable in these situations, but good luck actually trying to get them deployed, you're on your own". Each vendor does randomly different things for UI / credential management / workflow / whatever. The end result is that the spec is largely theoretical. In practice, people do any number of hacks to get something to work. Because the specs don't help here. If people can't deploy a spec easily and securely, then I see that as a failure of the specification. For example, over the last 20+ years, the "Security Considerations" section of RFCs has grown in importance and content. This is a good thing. > All of this having been said, perhaps the best way to go forward is to have a > requirements discussion in terms of the sorts of operations we would like to > see as part of the authentication process – as opposed to elsewhere. > > I see tremendous opportunity here, to be honest. But it's a lot of work. I agree. Alan DeKok. _______________________________________________ Emu mailing list Emu@ietf.org https://www.ietf.org/mailman/listinfo/emu
