You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As an alternative to wrapping the socket, we could define and implement the trait PathAwareDatagramSocket on UdpSocket and move the methods that are related to path persistence to it:
traitPathAwareDatagramSocket{fnpath();fnset_path();fnset_path_service(PathService);fnsend_to();fnsend();// Method to propagate paths from recvs to PathServicefnon_received_path();}
The upside: we would be including the delay in the destination AS in the measurements.
The downside: Measurements are done per-destination and so cannot be shared between different destinations in the same AS.
The pan implementation in golang seems to use Ping instead of traceroute to select paths. Do we want to use the same approach?
Yes, I became aware of that as well. The aliveness check (e.g., for scion showpaths) uses traceroute while the path selection in pan uses pings. In the current deployments, there is probably very little difference between the two in terms of measurement results as we usually have very small source and destination ASes.
I don't have a strong opinion but would probably go with the traceroute approach to start with.
- Always choose SCION if it is available?
- Compare performance?
Probably depends on #54
The text was updated successfully, but these errors were encountered: