Re: [icnrg] [EXT][Nfd-dev] about draft-irtf-icnrg-IPOC

[Junxiao Shi <[e-mail]>]

Hi Susmit


> But do we have to buffer the payload in the PIT? I know this is how this
> is done in NFD - but what do we lose if we throw away the payload (or
> application parameters) before populating the intermediate PITs? This might
> require changes in the forwarding pipeline.
>

If the forwarder doesn't store the whole Interest, its strategy cannot
retransmit the Interest except when processing an Interest or a Nack.

Also, a recent protocol clarification forces the forwarder to store the
entire Interest packet if it supports both PIT aggregation and Nack
returning.
https://redmine.named-data.net/issues/4535 note-16 requires the forwarder
to include the original Interest packet in a Nack returned to downstream.
To satisfy this requirement, the forwarder has to separately store the last
Interest from each downstream node. Sending the Nack from upstream to
downstream would not work, because the Interest enclosed in an upstream
Nack reflects the original Interest from at most one downstream node, while
other downstream nodes may have sent a different set of InterestLifetime,
HopLimit, as well as unrecognized non-critical fields.

NDN-Lite could get away with storing only the name and CanBePrefix flag
because it does not support Nack returning. The NDN protocol (
http://hdl.handle.net/10150/625652 chapter 3) defines Nack returning to be
optional, so this is a valid implementation choice, at the expense of
slower link failure recovery.

Yours, Junxiao

>