Falcosc
2017-02-21 18:23:03 UTC
Let me give you an example:
A swarm of 5 hosts, 1 seed and 4 peers.Seed has enable read cache suggestion.
What does happen?
All clients request the suggestion and then all clients have same pieces which means they can’t share there bandwith to speed up the swarm.
Could anybody confirm this behavior? Or does some clients ignore the suggestion in same cases?
The following could be a valid case to ignore a suggestion:
3 of 4 leecher does have a piece. The last leecher got a suggestion from seeder and discover that this suggestion is available by all peers. Now the leecher should ignore the suggestion and download something which is not available in swarm.
I would describe this logic like this:
Only use suggested pieces if they are missing in swarm, ignore suggestions which are available by all peers (definition of all peers is unclear because we should only consider active peers)
What do youthink about it?
Another way to work with that could be a recommendation in our documentation:
You should use suggest_read_cache only in combination with volatile_read_cache because otherwise leechers would be not able to share their bandwidth if all leechers get same piece suggestion.
Would be nice if you could tell me how the suggestions are handled by us and by other clients.
Kind Regards
Falco
A swarm of 5 hosts, 1 seed and 4 peers.Seed has enable read cache suggestion.
What does happen?
All clients request the suggestion and then all clients have same pieces which means they can’t share there bandwith to speed up the swarm.
Could anybody confirm this behavior? Or does some clients ignore the suggestion in same cases?
The following could be a valid case to ignore a suggestion:
3 of 4 leecher does have a piece. The last leecher got a suggestion from seeder and discover that this suggestion is available by all peers. Now the leecher should ignore the suggestion and download something which is not available in swarm.
I would describe this logic like this:
Only use suggested pieces if they are missing in swarm, ignore suggestions which are available by all peers (definition of all peers is unclear because we should only consider active peers)
What do youthink about it?
Another way to work with that could be a recommendation in our documentation:
You should use suggest_read_cache only in combination with volatile_read_cache because otherwise leechers would be not able to share their bandwidth if all leechers get same piece suggestion.
Would be nice if you could tell me how the suggestions are handled by us and by other clients.
Kind Regards
Falco