[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: request to recharter



Stefan Winter <mailto:stefan.winter@restena.lu> scribbled on Wednesday, February 13, 2008 9:45 PM:

> Hi,
>
>> Yes, there has been a popular upswell of support for abandoning our
>> work in favor of yours -- oh, wait, that would be you alone...
>
> well, first: it's not abandoning abandoning RADIUS work. In fact, the
> RADIUS payload is left completely untouched.
>
> Second: it wasn't me alone. I was the one who showed up at IETF,
> that's right.
> But as you very well know, there was an implementation out there for
> years which had customers, and RadSec has proved useful to a
> worldwide-operating roaming consortium, and has been implemented by
> other parties meanwhile. If I had gotten an outright reject from the
> radext group as whole I wouldn't have taken the step to ask for
> rechartering. But if you recall IETF69+70, there *was* support for
> the work. Yes, there has also been an outcry to ignore RadSec
> completely -- oh, wait, that was you alone... 

I'm afraid you are incorrect.  From your first appearance at radext I have expressed the opinion that, while the radsec work was interesting, it is not RADIUS & therefore does not belong in the radext WG. 
 
>
>> No, no, NO!  "radsec" is not RADIUS, period.  Not an extension, not
>> even a perversion, not RADIUS.
>
> We've had a discussion on that topic before on this list. That time,
> at the end you resorted to a strange analogy of ducks and snakes
> IIRC, not providing a reasonable argumentation. 

I assume that you are referring to the following exchange (https://ops.ietf.org/lists/radiusext/2007/msg00603.html):

<QUOTE>
> Well, except that in order to place these things in a common category it
> is necessary to ignore one of the defining characteristics of RADIUS:
> that it is only and intentionally defined as a connectionless protocol.

>> The impact of connectionless vs. connection is not very significant when
>> discussing the protocol's *security*, IMO.

So I gather that as far as you're concerned a network "protocol" solely consists of the format of the innermost PDU?

> might like to believe so, RadSec is _not_ RADIUS,

>> Well, it's _almost_ RADIUS, especially if one takes into account the
>> mechanisms in RADIUS to detect duplication, retransmission etc, which already
>> contains parts of the merits of TCP.

By that logic, a cobra is _almost_ a pigeon.

> & therefore far out of
> scope of not just the crypto-agility discussions but of this WG.
> RADIUSoDTLS, OTOH, doesn't alter the fundamental nature of RADIUS &
> therefore _is_ in scope.

>> I have understood that DTLS is considered in scope, while RadSec is out of
>> scope (Reminder: I never claimed otherwise). Just keep in mind that DTLS
>> itself requires even more characteristics of TCP, so when combining the
>> RADIUS mechanisms that come on top of UDP + the transport mechanisms in DTLS,
>> the end result is getting very close to TCP already. So the distance between
>> in-scope and out-of-scope is a very thin line.

Only if snakes are almost birds...
</QUOTE>

I'm sorry that you didn't understand that simile.  Perhaps you also don't recall (or didn't understand) the message that elicited it (https://ops.ietf.org/lists/radiusext/2007/msg00590.html):
<QUOTE>
> I agree that it is not relevant for crypto-agility discussions. As I
> see it, DTLS and RadSec fall under a common category in this respect:
> whole packet encryption, with encryption negotiation happening
> outside of RADIUS.  

Well, except that in order to place these things in a common category it
is necessary to ignore one of the defining characteristics of RADIUS:
that it is only and intentionally defined as a connectionless protocol.
This is not an accidental result of the transport choice as you seem to
imply below ("{some transport}+{the equivalent of TLS for that
transport}"), but a fundamental quality of RADIUS itself.  Much as you
might like to believe so, RadSec is _not_ RADIUS, & therefore far out of
scope of not just the crypto-agility discussions but of this WG.
RADIUSoDTLS, OTOH, doesn't alter the fundamental nature of RADIUS &
therefore _is_ in scope.
</QUOTE>

Since the "strange analogy" you deride was an attempt to explain the above statement (what part of it is "unreasonable", exactly?)  I won't risk further derision by attempting to explain it again.

>
>> I would support a radsec BOF & WG, but I will not support this.
>
> *Thank you*! It is very nice to read that you would support a WG for
> that work, which appears to imply that you see the concept as worthy
> for exploration and eventual standardisation.

See above; actually, I'm not really sure about standardization, but surely it could be documented. 

> Actually, I had been
> thinking about this way forward as well. The outcome was that
> creating a whole new working group for this one thing probably isn't
> justified. After all, it is not likely to create follow-up work, it's
> pretty much a standalone thing to add to RADIUS. So, if a working
> group isn't justified due to lack of amount of work, but the work is
> still somewhat important, where would the place to go be? My
> conclusion was that radext is the place to go.

As the chairs have pointed out, the WG has almost completed its work & it seems like the only thing being considered to continue radext is radsec.  So if there is not enough work to justify a WG (something I doubt very much) then it doesn't seem to justify the continuation of radext, either.  Just publish an Informational RFC & be done with it.  BTW, what do the members of the "worldwide-operating roaming consortium" think about giving up change control to the IETF (or do they think that we will simply rubberstamp this work of sheer genius)?

> And I still think that
> reasoning is justified, Bernard's reply in the thread looks like it
> is at least deemed worth a consideration.         
>
> Greetings,
>
> Stefan Winter