On Wed, 2006-05-31 at 10:08 +0530, Vishwas Manral wrote: > Hi Jeroen, > > You have raised some very valid points. I will add the proposal you > define, "e." to the draft. I also agree and will replace Issues with > issue e.g. "Issues with Policy Boxes" with "Issue with policy boxes" > > I am however not sure of the issue you raise with Section 6., as > bogus. The bogus part mostly because it affects already deployed IPv6 stacks and if they don't adhere to the specified behaviour would become broken if the wording like stated would be made mandatory. > I tried reading the text a couple of times, but could not > understand it. The proposal is to make the non-last fragement slightly > lesser than 1280, to address any fragmentation if tunnelling was to be > done. The -02 version reads in section 6: "To prevent such a case a minimum packet size of the non-last fragment should be less then 1280." Thus a non-last fragment should be minimally less than 1280? I guess you mean that it should be AT MOST 1280. But that would mean that anything above 1280 will never be used. This contradicts with the line before it: "The minimum fragment size of the non last fragment could be specified to be 1280 octets, the minimum link MTU [RFC2460]." Which could be a sane thing to specify and most likely what most implementations at the moment do, but it might be that some current implementations don't do this and have a small packet in the middle of the fragments and thus that would cause these implementations to break. > I do not feel it needs to be done that way by all implemntation, > but we can raise the concern that might arise if we make the fragment > size as 1280 or more. Am I missing the point altogather? No not missing the point, just miscommunication ;) Greets, Jeroen
Attachment:
signature.asc
Description: This is a digitally signed message part