So you are sending the ICMP error with inside the uncompressed packet? Maybe 
it’s ok for hop count but what if the error is due to malformed/ unsupported 
form of compression? The sender may not recognize what it sent in the attempt 
of decompression that is returned. When you detect the error did you already 
loose track of the compressed packet? I thought you’d keep it to forward it 
because most times there’s no need to change the iphc piece.


Regards,

Pascal

Le 6 nov. 2019 à 18:42, Martine Lenders <[email protected]> a écrit :


Hi Pascal,

Am Mi., 6. Nov. 2019 um 18:16 Uhr schrieb Pascal Thubert (pthubert) 
<[email protected]<mailto:[email protected]>>:
Hello Martine

You did great ! Yes I think that
You should send back the beginning of the original fragment at least up to the 
HL in an icon error and then an  abort...

This is the generic problem with errors in 6LoWPAN HC networks . Should the 
ICMO contain the compressed or uncompressed?
To sort out the error which maybe with the compression I believe we must send 
back the compressed form in ICMP errors .

That is contrary to what I did and also seems to me way more complicated than 
they need to because:
a) IPHC (which isn't currently "aware" of ICMPv6 as it is not compressed) needs 
to parse through every intermediate non-compressed header at *every* hop (as 
the compression changes at every hop).
b) If you have a classic IPv6 interface running in tandem (e.g. the border 
router, but we also have use cases were people just have an Ethernet cable on a 
6LN as well for diagnostics) you have to implement two kinds of ICMPv6 error 
messages: compressed and uncompressed.

So having them to have compressed (MUST) would add a lot of overhead.

Best regards,
Martine



Regards,

Pascal

> Le 5 nov. 2019 à 11:54, Martine Lenders 
> <[email protected]<mailto:[email protected]>> a écrit :
>
> 
> Hello,
>
> I'm aware this is a corner case that usually should not happen in a properly 
> configured WPAN, but what is an implementation of selective fragment recovery 
> supposed to do, if hop-limit 0 is reached? With minimal forwarding I just 
> reassembled it and handed it to IPv6 so it can send a ICMPv6 Time exceeded 
> error message.. Should selective fragment recovery send an abort ACK in that 
> case as well? Or just the abort ACK?
>
> Best regards,
> Martine
> _______________________________________________
> 6lo mailing list
> [email protected]<mailto:[email protected]>
> https://www.ietf.org/mailman/listinfo/6lo
_______________________________________________
6lo mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lo

Reply via email to