Hi everyone, (comments inline too) Am 29.01.2015 um 10:42 schrieb Victoria Mercieca <vmercieca0@xxxxxxxxx>: > Hi all, > > Comments on issue 56 and 57 in line... > > > On Tue, Jan 27, 2015 at 9:10 PM, Charlie Perkins > <charles.perkins@xxxxxxxxxxxxx> wrote: > Hello folks, > > I have gone through the above remaining issues. We need to discuss > issue #56. Stan was going to wordsmith something about how RFC 5444 > was in charge of security. We were all going to read RFC 7182. I did > start to read it. I need to provide text for issue #36, but I'd like to first > see what Stan's wordsmithing would look like. > > Most of the other remaining issues not mentioned below already have > some proposal for resolution. > > Can we have another conference call this Friday? > > Regards, > Charlie P. > > > =========================================================== > > === 36 Security Considerations: Reactive protocol concept === > > Need to update Applicability statement to explain when reactive > protocols are applicable even in the face of the threat as described. > > === 37 Security Considerations: what needs to be implemented? === > > Stan's solution, wordsmithed for this. > RFC 5444 problem. > > === 38 difficulty to do security, in case messages are mutable === > > Stan's solution, wordsmithed for this. > RFC 5444 problem. > > === 40 AckReq vs RREP_ACK === > > Reject with explanation that AckReq triggers RREP_ACK, so that > they serve different purposes. > > === 42 What happens if Active routes exceed RERR packet size? === > > Send multiple RERR packets in this case. Add sentence to draft. > Suggest closing issue. > > === 43 Reliance on bidirectional paths (submitted for Chris Dearlove) === > > We are now making RREP_ACK mandatory to implement. > Suggest closing issue. > > === 44 Hop count (submitted for Chris Dearlove) === > > Metric in AODVv2 is implementable by pure fact of being > required to be monotonically increasing. > Suggest closing issue. > > === 45 RFC 5498 non-compliance (submitted for Chris Dearlove) === > > Areas of noncompliance have been revised. > Suggest closing issue. > > === 47 Security approach unacceptable (submitted for Chris Dearlove) === > > Stan's solution, wordsmithed for this. > RFC 5444 problem. > > === 49 Locating pseudocode in appendix (submitted for Chris Dearlove) === > > Non-normative example text for specific cases provided as > a hint for implementers. > Suggest closing issue. > > === 50 Weak gateway support (submitted for Chris Dearlove) === > > Full gateway support is out of scope for the document. > Refer to previous discussion on mailing list. > Suggest closing issue, or else chartering a gateway draft. > Note: We had a gateway draft already that was implemented > and known working. > > === 56 Issue concerning RREQ redundancy check methodology and order === > > Look at moving section 7.5.1 so that RREQ redundancy precedes route table > check > > Check RREQ redundancy (using RREQ table): > - same metric type > - same OrigAddr and TargAddr > > Check route table (using route table): > - same metric type > - same TargAddr > - not stale SeqNum > - if SeqNum ==, then look at metric value, etc. > > > > 1) Following this scheme, a RREQ that is “redundant”, but with a > > better metric, would update the routing table, but *not* the > > RREQ table, since the RREQ Table only checks for metric Type, > > but not for metric value. In consequence, it would also not lead > > to a rebroadcasting of the new, better (in terms of metric) RREQ. > > This doesn't make sense to me, am I missing something here or is > > this in fact a defect? > > > The redundant RREQ typically does not have to be rebroadcast, since > any RREP will take the improved route back to OrigAddr. However, > RREP might not reach the intermediate node at all if some other route > looked better because of the nonoptimal metric back to OrigAddr. We > *could* specify that RREQs may be re-broadcast and add Metric to the > RREQ redundancy table. > > TO DISCUSS. > > I like this idea. The way I read the current draft, it already does check > metric to determine if the incoming RREQ should replace the one in the table. > I thought this would result in the RREQ being regenerated, but you are right, > if we are concerned about the metric being correct then it will need to be > re-broadcast, but the route should already exist because of the previous RREQ > so we could make it optional to re-send the second RREQ. ...But the already existing route could be a possibly worse one over an entirely different route, right? In any case, I like your suggestion. > > > > In case a), since the check for RREQ redundancy is performed after > > the routing table update, redundant RREQs may “update” the routing > > table with the exact same data it held before. This seems like > > computational overhead to me which could be solved by first checking > > the RREQ for redundancy and only updating the Routing table if this > > check turned out negative. However, this solution would only work > > if the RREQ table checked not only metric type, but metric value, > > as suggested in 1). Another solution would be to keep the order > > (routing table update, then RREQ table redundancy check) only update > > the routing table in case b). This would still involve an extra call > > to the routing table, but saves the effort of copying redundant data. > > Let's consider specifying that the RREQ redundancy check happens before > the route table check. > > TO DISCUSS. > > I think this should happen - with the check on metric in the RREQ. If the > incoming RREQ updates the RREQ table, it should then get processed as normal > (but maybe with the option to restrict re-broadcast). > +1 > === 57 Need to further restrict "LoopFree" condition === > > Fixed. > Suggest closing issue. > Or, wait for further analysis from Vicki. > > > I would like to analyse this a bit further. I have one example of where > removing the +1 that was previously part of the LoopFree condition results in > a route being ignored when it does not form a loop, and I have an idea for > the LoopFree calculation (which I've mentioned in my "Notes on AODVv2 Draft" > document under 6.5.2) but I'd like to test it in a few example scenarios and > maybe add an explanation before we close this issue. Would you be able to > send me the proof that Kedar Namjoshi and Richard Trefler provided for their > work on the LoopFree condition? > Will you be attending the call on friday too? If so, we could discuss this in depth... Cheers, Lotte > Regards, > Vicky. > > 61 Difference between "broken" and "expired" (Submitted for Justin Dean) > > Propose to coalesce the two states. My analysis shows > no harm, resulting improvement from simplification. > Suggest to list and closing the issue.