Re: [6lo] draft-ietf-6lo-fragment-recovery: Send a FULL bitmap when datagram is complete?

["Pascal Thubert (pthubert)" <[e-mail]>]

Hello Martine

I meant fragments. It can be expected that a few fragments are still in flight after everything is received because of end to end fragment retries or L2 ARQ. So the receiver must keep a state to drop them silently. But is it gets “too much “ of that it may be an error and the receiver should abort the flow.

That “too much” decision is left to implementation.


Regards,

Pascal

Le 23 oct. 2019 à 14:29, Martine Lenders <[e-mail]> a écrit :


Hi Pascal,

Am Mo., 21. Okt. 2019 um 18:14 Uhr schrieb Pascal Thubert (pthubert) <[e-mail]<mailto:[e-mail]>>:
Hello Again

I reread the text and it appears that the receiver operation is too implicit. I suggest to add this in the last fragment processing:

    When all the fragments are received, the receiving endpoint reconstructs
    the packet, passes it to the upper layer, sends a RFRAG Acknowledgment on
    the reverse path with a FULL bitmap, and harms a short timer to absorb
    packets that are still in flight for that datagram without creating a new
    state and abort the communication if it keeps going.

Does that help?

If this goes somewhere in section 6, yes I think that makes it far more understandable.

Note that there’s room for an implementation to decide if it absorbs silently a few packets and for how long, and when it decides to reset the flow. The all 1 (to be renamed throughout to FULL)  does not help more than the reset.

By packets you mean fragments or reassembled datagrams. I don't really understand what you mean by that.

Best regards,
Martine


Pascal


From: Pascal Thubert (pthubert)
Sent: lundi 21 octobre 2019 17:29
To: Martine Lenders <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: RE: [6lo] draft-ietf-6lo-fragment-recovery: Send a FULL bitmap when datagram is complete?

Sorry I missed that Martine!

The ALL 1s was already sent when the last fragment was received. This text happens later.

It is supposed to have been processed along the way back. The receiving end node maintains a state for a “short” time after the message processing to absorb packets that may still be in flight. During that “short” time it is capable to recognize redundant packets and drop them as opposed to create a new state and expect the full fragment. For legitimate packets still in flight the good thing would be to stay silent. If the Ack with a FULL (All 1s) bitmap was lost then sending it again would be OK as you point out.

But there might also be error conditions, like a weird situation that the FULL bitmap did not fix on its way back where the sender keeps sending. If the FULL bitmap failed then retrying it may fail again. The reset is a clearer indication to drop everything regardless and move to the next.

Works? Should we massage text?

All the best

Pascal

Am Di., 1. Okt. 2019 um 16:31 Uhr schrieb Martine Lenders <[e-mail]<mailto:[e-mail]>>:
Hi,

draft-ietf-6lo-fragment-recovery states in section 6.3

[the] might need to abort the process of a fragmented packet for internal reasons, for instance if it […] considers that this packet is already fully reassembled and passed to the upper layer. In that case, the receiver SHOULD indicate so to the sender with a NULL bitmap in a RFRAG Acknowledgment.

The given example seems to me the perfect instance to set a FULL bitmap instead. There is no other instance were a FULL bitmap is specified to be sent, except for the case that the datagram incidentally fills out the whole value space of the sequence number field.

Or am I missing something?

Kind regards,
Martine