[Sis-csi] Text for Transport Layer Section

[Scott Burleigh]

Lloyd Wood wrote:

> But it's interesting that, as you say, CFDP was designed for a single 
> 'one-hop' radio link, effectively treating IP as a link protocol,

That's right.  More precisely, treating the "UT layer" protocol stack -- 
whatever it is: CCSDS space packets over AOS frames, UDP datagrams over 
IP packets over Ethernet (or whatever), etc. -- as a link protocol.

> and assuming an ordered link flow. (How that ordered flow is produced 
> or ensured can vary.)

Yes, exactly.

> Analogous to TCP, which suffers in out-of-order delivery on its path. 
> The designs have to be robust against out-of-order-delivery, but still 
> treat it as a rare edge case that doesn't need to be optimised for.

Yes.  CFDP is designed not to break in any major way if data arrive out 
of transmission order, but it's optimized for the expected case of 
in-transmission-order arrival.  And there are warnings sprinkled all 
through the spec about acknowledged-mode features that you especially 
shouldn't rely on if out-of-order arrival is potentially frequent.

Another point that might be worth mentioning: the recently added 
"Unacknowledged CFDP Extensions" improve CFDP's tolerance of 
out-of-order arrival in Unacknowledged mode a bit further, aiming to 
support deployments where a reliable UT layer protocol underneath 
unacknowledged CFDP is retransmitting lost data (the retransmitted CFDP 
PDUs are bound to be out of the original file segment transmission order).

Scott