Re: [6lo] recoverable fragments: should we adapt the fragment size to ECN?

On Tue, May 21, 2019 at 3:05 PM Pascal Thubert (pthubert) <
pthubert@cisco.com> wrote:

> Dear all:
>
> *   During the review by Laurent, a question came up on we could change
> the fragment size when transmitting a datagram. As it goes we can, because
> the offset and size are indicated in each fragment. If the MTU goes down
> and a fragment must be resend, we can also resend a smaller piece with the
> original fragment sequence and send other fragments with unused sequences
> for the rest of the original fragment. This is made possible because the
> sequence is sequential in the bitmap but not in the order of fragments.
> That being said, a reaction to an ECN is to reduce the window. At the same
> time a node could also reduce the datagram size and enable a more
> streamlined transmission for all.   What do you all think? *
>

I think we should not change size to smaller when transmitting, but if
needed it must be limited for special cases.

AB

> *  Pascal   *
>
>
>
>
> * _______________________________________________ 6lo mailing list
> 6lo@ietf.org <6lo@ietf.org> https://www.ietf.org/mailman/listinfo/6lo
> <https://www.ietf.org/mailman/listinfo/6lo> *