[Nfd-dev] A Potential Conflict in Developer Guide about FIB Entry's NextHop Collection

[Ju Pan]

Hi Junxiao,

Thanks for your clarification.

So, for FIB entry with a non-zero length name, it's supposed to have a
non-empty nexthop collection right?

Please forgive me bringing up a separate issue here. In issue 4931
(aka async-strategy) (https://redmine.named-data.net/issues/4931), If
there is already a nexthop in the FIB entry (I guess the FIB entry
here should have a non-empty name?), shouldn't we forward the Interest
directly to that nexthop instead of retaining it? So based on my
understanding, the issue should be "Transmit pending Interests upon
FIB entry insertion". That is, we retain the PIT entry only when there
is no FIB entry matching.

Correct me if I was wrong. Thanks.

Regards,
Ju

On Thu, Jul 4, 2019 at 3:11 PM Junxiao Shi <shijunxiao at email.arizona.edu> wrote:
>
> Hi Ju
>
> FIB entry with zero length name may have zero next hop records. Such an empty isn't stored in FIB, but FIB lookup methods may return it so that forwarding doesn't need to check whether FIB entry is null.
>
> Yours, Junxiao
>
>
> On Thu, Jul 4, 2019, 17:46 Ju Pan <pjokk722 at email.arizona.edu> wrote:
>>
>> Dear folks,
>>
>> In NFD Developer Guide, there is a potential conflict in the
>> description about a Fib entry's nexthop collection.
>>
>> 1. In Chapter3 Tables --> 3.1 Forwarding Information Base (FIB) -->
>> 3.1.1 Structure and Semantics
>> --> FIB entry and NextHop record, it says:
>>
>> "A FIB entry (nfd::fib::Entry) contains a name prefix and a non-empty
>> collection of NextHop records."
>> (It says a FIB entry contains a non-empty collection of NextHop
>> records, in 3.1.2 Usage part, it says a FIB entry losing its last
>> nexthop record is deleted.)
>>
>> 2. In Chapter 4 Forwarding --> 4.5 Helper Algorithms --> 4.5.1 FIB
>> lookup --> 1, it says:
>>
>> "FIB guarantees that Longest Prefix Match returns a valid FIB entry;
>> however, a FIB entry may contain empty set of NextHop records, which
>> could effectively result (but, strictly speaking, is not required to
>> happen) in the strategy rejecting the Interest."
>> (It says a FIB entry may contain an empty set of NextHop records.)
>>
>> I think the actual NFD was implemented as the former way. (erase the
>> Fib entry if it doesn't have any nexthop)
>>
>> Could someone confirm this conflict or point out what did I
>> misunderstand? Thanks.
>>
>> Regard,
>> Ju
>>