Re: [v6ops] Operational Implications of IPv6 Packets with Extension Headers

[tom petch <[e-mail]>]

From: Fernando Gont <[e-mail]>
Sent: 15 September 2020 12:24

Hi, Tom,

On 15/9/20 06:05, tom petch wrote:
[...]
>
> I wouldn't mind writing a Section that sits between the current Sections
> 2 and 3 with more background on extension headers, if you think that
> would be of value.
>
> <tp>
> Yes please.  One page or two pages max, summarising why they were thought a good idea, the different types and their uses, how widely used they are  and with references;  RFC8200 may be Normative but it takes a while to appear and that after RFC2460!

An intro to EHs -- and e.g., how the structure compares to the IPv4
packet structure is simply and doable. OTOH, what you suggest might now. :-)

e.g.,Not sure one can tell how widely used they are. Also, getting into
enumerating EHs will result in the relevant section becoming stale soon,
since there seems to be quite a bit of development in this area.

<tp>
My sense is of the middle part of the I-D, the technical detail, is floating in mid-air, lacking an introduction, lacking a conclusion. and as such will only convince those who are already convinced.  I have to read between the lines to divine that there is a connection between fragmentation and EH, between length of packet header and EH, between security and EH and so on.  By not making the connection explicit, I think that you will lose your target audience.

Tom Petch


> Section 3 makes reference to security, fragmentation and such like with no apparent connection to EH; I think that such a connection should be explicit earlier.

I'll craft some text and post to the list for review before
incorporating into the I-D...


[...]
> <tp>
> but it leaves the I-D without an ending, it just peters out.  With an introduction saying what EH are used for then you could have a summary harking back to that and saying which, or which kind of, work and which do not implying, without being explicit, what future developments might or might not be a problem.

The think is that, on the public Internet, all that I have checked are
unreliable. -- That's a sad fact... but a fact. :-)

If you control the network, well.. you control whether they are dropped.

And the obvious corollary is that use cases that are targetted at
limited domains will likely work, where any targetted at the public
Internet will fail, at least for some time.

That's as much as we can say so far -- based on data... since otherwise,
guesswork is just guesswork.

Thoughts?

Thanks,
--
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492