I don't disagree with your latest statements. Those are generally
the assumptions I am working with. What I don't understand is the
link between those statements and your previous statements. The
current EAP authentication model is the default mode of operation if
something fails or if the new capabilities are not supported. By
that logic there shouldn't really be a blanket statement say one MUST
"use" EMSK. Of course, some access technologies may decide to
specify MUST try/use EMSK in the first round or something like that,
but that is not going to be in an RFC.
Anyway, that's my current thinking. We are early in the process here
and when the design is complete, we can think further about what makes sense.
In the meanwhile, let's stick to what we have in other RFCs.
Happy thanksgiving weekend to all
Lakshminath
At 06:07 AM 11/22/2006, Yoshihiro Ohba wrote:
On Wed, Nov 22, 2006 at 09:35:46AM -0800, Lakshminath Dondeti wrote:
> At 03:46 AM 11/22/2006, Yoshihiro Ohba wrote:
> >Hi Uri,
> >
> >On Wed, Nov 22, 2006 at 09:41:03AM -0500, Blumenthal, Uri wrote:
> >> >The problem we are facing with EMSK is that the key and its usage were
> >> >not defined *at one time".
> >>
> >> True, definition of EMSK has been done in advance, PRIOR to its usage -
> >> so that by the time EMSK usage is defined, all the compliant methods
> >> already have implemented EMSK generation and have it available. This is
> >> not a problem.
> >>
> >> Especially since it is likely that most real-life implementations are
> >> modular enough to allow component upgrades.
> >
> >For any method, how a compliant implementation and a non-compliant
> >implementation can be distinguished? If they are indistinguishable,
> >then there will be an interoperability problem.
>
> This is not rocket science! I have some second-hand experience with
> IPsec interoperability; if implementation X does not interoperate
> with implementation Y, people then figure out which one is
> non-compliant and have that implementation fixed.
A HOKEY compliant implementation should be able to revert to the
legacy mode of operation (i.e., use of MSK in the current way) so that
it still interoperates with a HOKEY non-compliant implementation
without ending up with a communication failure or changing the HOKEY
non-compliant implementation. Note that a HOKEY non-compliant
implementation is still RFC3748 compliant as long as it generates MSK
and EMSK even if it does not use EMSK.
>
>
> >For a future method that will be defined after we define an EMSK
> >usage, I agree that we can definetly mandate use of EMSK.
>
> This is moot. EMSK derivation is already defined as mandatory. Are
> you proposing something stronger than a MUST?
"EMSK derivation is a MUST" does not mean that "use of EMSK is MUST".
Yoshihiro Ohba
>
> Lakshminath
>
> >But
> >ignoring the existing implementations does not make sense, considering
> >the fact that some of them are already part of Wi-Fi certificate.
> >
> >Yoshihiro Ohba
> >
> >
> >>
> >> The problem is shortsightedness of some implementors, whose logic is "if
> >> I don't need this feature NOW, no way I'll spend resources implementing
> >> it".
> >>
> >> >For this reason, I don't agree with
> >> >mandating the use of EMSK for all implementations.
> >>
> >> We cannot mandate what has not been defined yet. This is why we have
> >> mandated only EMSK generation, not EMSK usage.
> >>
> >> Note the grammatic form of the verb: it IS a done deal, RFC 3578 IS a
> >> standard.
> >>
> >> > Only making use of
> >> >EMSK optional makes sense to me, regardless of how it is going to be
> >> >used.
> >>
> >> There is no use of EMSK currently defined, neither optional nor
> >> mandatory. So no need to make it anything.
> >>
> >> It is mandatory to generate EMSK and make it available for applications.
> >>
> >> How many times does is this dead horse going to be beaten?
> >>
> >>
> >> Yoshihiro Ohba
> >>
> >> On Tue, Nov 21, 2006 at 08:50:46AM -0800, Madjid Nakhjiri wrote:
> >> > Hi Yoshi,
> >> >
> >> > I don't quite agree with this. Defining the full functionality of a
> >> new EAP
> >> > method (including how it spits out MSK and EMSK) should be the job the
> >> EAP
> >> > method designer. 3748, EAP keying framework and hokey specs then
> >> define how
> >> > MSK or EMSK is used. This is modular design. People who use EMSK then
> >> do not
> >> > have to worry about the crypto behind EMSK generation.
> >> > Using your analogy is "go buy the key and lock" but then wait to see
> >> which
> >> > door you want to install it on:)
> >>
> >> > I as a home owner do not need to know how to make locks.
> >> >
> >> > R,
> >> >
> >> > Madjid
> >> >
> >> > -----Original Message-----
> >> > From: Yoshihiro Ohba [mailto:[EMAIL PROTECTED]
> >> > Sent: Friday, November 17, 2006 11:36 AM
> >> > To: Blumenthal, Uri
> >> > Cc: [EMAIL PROTECTED]; [email protected]
> >> > Subject: Re: [Hokeyp] [Emu] Re: MSK but no EMSK
> >> >
> >> > Hi Uri,
> >> >
> >> > I have a different view. Mandating generation of EMSK without having
> >> > defined its usage *when it was introduced* seems not much different
> >> > from not having defining it at all. It looks like selling a key
> >> > without a lock, make the lock later and say "You MUST use this key and
> >> > lock for your car." Make sense??
> >> >
> >> > Yoshihiro Ohba
> >> >
> >> >
> >> >
> >> > On Fri, Nov 17, 2006 at 09:22:04AM -0500, Blumenthal, Uri wrote:
> >> > > >The discussion focuses on the problem EMSK is optional or
> >> mandatory.
> >> > >
> >> > > I don't think this is a problem - GENERATION of EMSK is compulsoty
> >> as
> >> > > spelled out in RFC 3578.
> >> > >
> >> > > The problem is non-compliance. Some, er, people seem to think "the
> >> > > standard says do A, but since I don't use A at the moment - I won't
> >> > > bother."
> >> > >
> >> > > >RFC3578 defined EMSK is mandatory,
> >> > >
> >> > > And that should be the end of discussion.
> >> > >
> >> > > > but it is not used at all.
> >> > >
> >> > > First - do you know all the applications that use key-generating EAP
> >> > > methods? But really - who cares?
> >> > >
> >> > > >If EMSK must be used, it is mandatory. if no, I think,
> >> > > >it may be better that it is optional.
> >> > >
> >> > > VERY strongly disagree. Mandatory is what is explicitly specified as
> >> > > mandatory, period. Otherwise many would implement just those pieces
> >> and
> >> > > features of the standard that his particular product needs today.
> >> > >
> >> > > (I'm proud of my restraint - not even once using a term "B*S*" :-)
> >> > > _______________________________________________
> >> > > Hokeyp mailing list
> >> > > [EMAIL PROTECTED]
> >> > > http://www.opendiameter.org/mailman/listinfo/hokeyp
> >> > >
> >> >
> >> > _______________________________________________
> >> > Emu mailing list
> >> > [email protected]
> >> > https://www1.ietf.org/mailman/listinfo/emu
> >> >
> >> >
> >>
> >>
> >
> >_______________________________________________
> >Emu mailing list
> >[email protected]
> >https://www1.ietf.org/mailman/listinfo/emu
>
>
_______________________________________________
Emu mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/emu
