NFD

I asked @Wentao Shang for Selector usage in BMS application.
He mentions a use case:

• Producer names Data with a timestamp NameComponent
• Consumer expresses Interests to request Data sequentially. Every Interest has the same Name but different Selectors. To request Data in time period (T1,T2), the consumer would:
  1. request with Exclude=Any,T1
  2. when this request is answered, suppose the timestamp in Data is T3, the next Interest would be sent immediately with Exclude=Any,T3
  3. repeat until the timestamp in Data >= T2
• Consumer sends the next Interest as soon as the previous one is answered. If network latency is low, Interest rate can be very high.

In this use case, it shall be beneficial to organize PIT entries in the same NameTree entry into a hashtable, instead of the current doubly linked list.

I want to clarify the needed profiling as mentioned in note-4:

• INPUT: Interests with same Name, but many different Selectors
• find out whether std::list m_pitEntries in NameTree entry is a bottleneck, and whether changing it to std::unordered_map would improve performance

The performance of evaluating each individual Selector is irrelevant to this Task.

I have seen the course project report. It doesn't have the benchmark required by note-4 and note-7.

@Wentao Shang, please do the profiling as you promised in note-6.

At 20151110 conference call, Wentao claims that he will never have time to do the benchmark promised in note-6.

Lixia suggests that we can reject this issue, because there's no evidence it's needed.

20151117 conference call decides that we won't need this for now.