Re: [tcpm] I-D Action: draft-ietf-tcpm-tcp-edo-02.txt

[Joe Touch <[e-mail]>]

On 4/15/2015 11:45 PM, Yoshifumi Nishida wrote:
> Hi,
> 
> On Wed, Apr 15, 2015 at 8:59 PM, Wesley Eddy <[e-mail]
> <mailto:[e-mail]>> wrote:
> 
>     On 4/15/2015 2:10 PM, Joe Touch wrote:
>     > It includes a variant that indicates the TCP segment length to detect
>     > when coalescing that might interfere with EDO occurs.
> 
>     This is one thing that I think we particularly want feedback from
>     the group on.
> 
>     I think it's a very low-impact way (compared to adding checksums,
>     using AO, using IPsec, etc) in order to reasonably detect an
>     accidental pollution of the user data due to resegmentation.
> 
> 
>>> When an endpoint receives a segment using the 6-byte EDO
>    Extension option, it MUST validate the Segment_Length field with the
>    length of the segment as indicated in the TCP pseudoheader. If the
>    segment lengths do not match, the segment MUST be discarded and an
>    error SHOULD be logged in a rate-limited manner.
> 
> 
> I am wondering if discarding a segment is safe enough.
> Once resegmentation happens, I think we are not very sure how to recover it.
> I guess it would be better to reset a connection.

It's safe - the connection will fail after several failed
retransmissions anyway. That limit can be controlled with the TCP User
Timeout option.

Jumping to a reset would be quicker but might be premature. The sender
might be able to resend segments without EDO or the situation might be
transient.

Joe