> joel jaeggli <mailto:joe...@bogus.com> > 10 June 2013 22:04 > On 6/10/13 9:35 PM, Ray Hunter wrote: >>> Christopher Morrow <mailto:christopher.mor...@gmail.com> >>> 10 June 2013 20:59 >>> On Mon, Jun 10, 2013 at 2:44 PM, Ray Hunter <v6...@globis.net> wrote: >>>>> Christopher Morrow <mailto:christopher.mor...@gmail.com> >>>>> 10 June 2013 17:22 >>>>> On Mon, Jun 10, 2013 at 10:56 AM, Nalini Elkins >>>>> >>>>> Some of the discussion already had talks about ordering and optimum >>>>> method to find X in the header chain. What happens in these >>>>> situations >>>>> when someone sends 'lots' of packets with 'bad ordering' of the >>>>> header >>>>> bits? Will the devices in the path behave 'well'? or will I get >>>>> degraded forwarding performance because of bad ordering? or will the >>>>> packets be dropped? or ? >>>>> >>>>> it's worth thinking about what can go wrong with this requirement to >>>>> order EH bits in certain ways. >>>>> >>>>> -chris >>>>> >>>> I purposefully left that particular can unopened, because I >>>> consider it >>>> unlikely that we will obtain rough consensus on it. >>> :) >>> >>>> IMHO it's better to have an informational RFC that says "if you obey >>>> these simple formatting rules, your packet is likely to be transported >>>> (fast)" than nothing at all. >>> right, but you're also asking HW manufacturers to optomize on a >>> certain ordering (sort of), that's going to lead to: 1, 2, 3 ordering >>> works great! >> Yes. Better than today. >> >>> but: 3,2,1 == slow-path :( >>> >>> or COULD lead to that. >> Same as today. So they'll continue to make their own decision on whether >> to deploy slow path or drop. > Understand that there is no slow path in high capacity devices. > > There is fast path or no path. > > If slow path exists it becomes a bottleneck or a dos target. > I understand exactly.
Which is why I am studiously ignoring the issue. :) regards, RayH >> >>>> I just searched draft-ietf-v6ops-6204bis-12 for requirements on >>>> extension headers. Nada AFAICS. >>>> >>>> I also just tested my own home CPE for outbound transmission of >>>> extension headers (with and without firewall configured): >>>> HBH EH: tick >>>> Destination Options EH: tick >>>> Fragment EH: looks like they're always dropped >>> neat... how about proto-59? (noext-headers)? :) >> Untested (no test source yet) >> >> regards, >> RayH >> > > -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
