At Sun, 05 Aug 2007 11:27:01 +1000, Geoff Huston wrote:
>
> Rob Austein wrote:
> > Having thought about this a bit, I remain skeptical about the need for
> > or desirability of multiple signatures on ROAs.
> > 
> > First, as others have mentioned, this is a relatively low-probability
> > hypothetical case, and if it occurs at all it would be the result of
> > an issuer deliberately chosing to make life complicated for its
> > subjects.  This does not strike me as a strong case for complicating
> > the protocol (if anything, it strikes me as the opposite, absent proof
> > that this complexity really is necessary).

I still stand by this part of what I said.

> > Second, I don't see why this can't be handled via multiple ROAs
> > instead of multiple signatures on a single ROA.  As I understand it,
> > relying parties in this system are going to have to deal with the
> > possibility of multiple ROAs for the same AS number in any case;
> > adding multiple signatures to ROAs will not change that.
> 
> So if seems to me that you are saying that an advertisement for 
> 192.0.2.0/24 originated from AS65000 could be validated by two ROAs, 
> namely 192.0.2.0/25 authorizing AS65000 and 192.0.2.128/25 authorizing 
> AS65000
> 
> To me, this appears to make the relying party's job harder given that 
> the relying party is no longer just looking for either exact match ROAs 
> or covering aggregate ROAs but now also has to search for a collection 
> of more specific ROAs that could be used to construct an aggregate that 
> matches the prefix to be validated.

Upon further consideration and some offline discussion, I agree with
you on this point.  I was confused about the other cases where I
thought one might need to construct an aggregate from multiple ROAs.

But this was a side point.  That my proposed work-around to a
non-problem is itself unworkable does not change the nature of the
non-problem.

> > So, since on the one hand this whole mess can be avoided by an issuer
> > who wants to avoid it, and on the other hand there's a perfectly good
> > way to handle it that we're going to have to support anyway, on the
> > gripping hand I do not support the proposed change to allow multiple
> > signatures.
> >
> 
> I don't agree with this assessment and to me ruling out the ability of 
> multiple signatures on a ROA introduces the potential for undue levels 
> of uncertainty in relying party validation of route objects. For that 
> reason I continue to support the concept of allowing multiple signatures 
> on a ROAs.

I don't see how this follows.  I still don't believe that there's any
real need for multiple signatures, because there's no good reason for
the issuer to force this mess upon its subjects.  As far as I can tell
this is a non-problem to which we do not need a solution.

_______________________________________________
Sidr mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/sidr

Reply via email to