fixed compilation of network simulator

git-svn-id: http://svn.code.sf.net/p/retroshare/code/trunk@7665 b45a01b8-16f6-495d-af2f-9b41ad6348cc

libretroshare/src/tests/network_simulator/README.txt

@@ -1,3 +1,51 @@
+TODO
+====
+
+Testing the router algorithm with network simulator
+	* generate a random network
+	* simulate disconnections (also in GUI)
+	* send messages from/to random peers and measure:
+		- how many times a given message is stored
+		- how much time a given message took to arrive.
+
+Implement
+	* clueing of GR from GXS (simulated in network simulator, when initing the network)
+
+In GLobal Router, by order of priority
+	* when a ACK is received for a msg that is already ACKed, we should still update the routing matrix and add a clue, but with lower prioity,
+		so that the matrix gets filled with additional routes. Can be checked in the simulator
+
+	* routing strategy:
+		- when a route is known and available, always select it, but possibly add another random route, very rarely.
+			Peer disconnection is likely to cause the discovery of new routes anyway.
+
+	* we should use clues from GXS to improve the routing matrix
+		- That would avoid lots of spamming.
+		- allows to init the routing matrices for all keys
+		- random walk will be a supplemental help, but restricted to small depth if no indication of route is available.
+		- needs to be implemented in network simulator. When providing a new key, the key should be spread in the network and new clues
+			should be added to the RGrouter matrix.
+
+	* make sure the depth is accounted better:
+		- if single route is known => don't limit depth
+		- if no route is known => strictly limit depth
+			=> add a counter which is increased when no route is available, and *reset* otherwise, so that the max number of bounce
+				we can do without knowledge of the keys is limited, but the total depth has no limits.
+
+Unsolved questions:
+	* handle dead routes correctly. How?
+	* should we send ACKs everywhere even upward? No, if we severely limit the depth of random walk.
+	* better distribute routing events, so that the matrix gets filled better?
+	* find a strategy to avoid storing too many items
+	* how to handle cases where a ACK cannot be sent back? The previous peer is going to try indefinitly?
+		=> the ACK will be automatically collected by another route!
+	* how to make sure ACKed messages are not stored any longer than necessary?
+	* send signed ACKs, so that the receiver cannot be spoofed.
+	* only ACK when the message was properly received by the client service. No ACK if the client does not register that item?
+
+================================================================================================================
+
+
 The purpose of this directory is to write a Network simulator, that can have multiple turtle routers interact
 together. The routers will talk to a fake link manager, which reports the peers for each node of a network graph.
 

libretroshare/src/tests/network_simulator/gui/gui.pro

@@ -20,7 +20,6 @@ FORMS = NetworkSimulatorGUI.ui TurtleRouterStatistics.ui GlobalRouterStatistics.
 
 LIBS *= ../../../lib/libretroshare.a \
         ../../../../../libbitdht/src/lib/libbitdht.a \
-		  ../../../../../../lib/sqlcipher/.libs/libsqlcipher.a   \
 		  ../../../../../openpgpsdk/src/lib/libops.a \
 		  ../lib/libnscore.a \
-		  -lgnome-keyring -lupnp -lssl -lcrypto -lbz2 -lixml
+		  -lsqlcipher -lgnome-keyring -lupnp -lssl -lcrypto -lbz2 -lixml

libretroshare/src/tests/scripts/config-linux.mk

@@ -71,7 +71,7 @@ LIBS +=  -lssl -lcrypto  -lpthread
 #LIBS +=  -L$(UPNPC_DIR) -lminiupnpc
 LIBS +=  $(XLIB) -ldl -lz 
 LIBS +=  -lupnp 
-LIBS += ../../../../../lib/sqlcipher/.libs/libsqlcipher.a
+LIBS += -lsqlcipher
 	
 RSLIBS = $(LIBS)