On 29/2/20 23:19, Joseph Touch wrote:
On Feb 29, 2020, at 5:46 PM, Fernando Gont <fernando@xxxxxxxxxxx
<mailto:fernando@xxxxxxxxxxx>> wrote:
I did look at the protocols involved here; the ingress does add
headers but doesn’t appear to handle fragmentation.
That’s a non-starter if you want your packets to traverse a network
because people WILL hand you 1280-byte packets, so what will you do?
FWIW, we have been insisting on this point (and others) since they
first tried to push EH insertion in
draft-ietf-6man-segment-routing-header.
THey removed it from *that* document, but they keep trying to push
similar ideas in other documents.
Well its seems simple to me - they need a plan for fragmentation or it’s
simply a nonstarter because they can’t support 1280-B packets traversing
the network.
Indeed.
But it's worse than that. The proponents, and others, claim that IPv6
supports extension header insertion/removal en-route to a destination.
In order for that to be the case, the IPv6 specification itself
(RFC8200) would have a plan to make PMTUD work (and IPsec AH, too). But
obviously it doesn't (because it never supporte EH insertion/removal
en-route to destination).
Still, the errata I submitted
(https://www.rfc-editor.org/errata/eid5933), is unprocessed, and at
least there was a statement that would be "held for document update"
(see:
https://mailarchive.ietf.org/arch/msg/ipv6/yVKxBF3VnJQkIRuM8lgWN4_G3-o/)
No amount of “but this is what the user wants” translates to “they want
their packets dropped silently”.
It smells even worse when they pretend that the IPv6 allows for their
proposed behavior.
Thanks!
Cheers,
--
Fernando Gont
SI6 Networks
e-mail: fgont@xxxxxxxxxxxxxxx
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492