NLSR

When a FIB entry is installed or updated, we schedule a refresh after FIB entry refresh time (which is 2 * LSA refresh time by default).
If the neighbor goes down and recovery is not done in 2*LSA refresh time then, the refresh is cancelled. So NLSR will no longer have a route to send the INFO interest to its neighbor.

To reproduce, run Mini-NDN mcn failure experiment with LSA refresh time: 240 seconds added to nlsr.py. Convergence time of 480 seconds with the following topology:

a----b

Scenario is explained in the attached figure.

Seems like the solution is to not schedule the refresh (register) of neighbor's router FIB entry:

@@ -142,7 +142,7 @@ Fib::update(const ndn::Name& name, const NexthopList& allHops)
     entryIt = m_table.find(name);

   }
-  if (entryIt != m_table.end() && !entryIt->second.getRefreshEventId()) {
+  if (entryIt != m_table.end() && !entryIt->second.getRefreshEventId() && isPrefixUpdatable(entryIt->second.getName())) {
     scheduleEntryRefresh(entryIt->second,
                          [this] (FibEntry& entry) {
                            scheduleLoop(entry);