While auditing the diffusion/peer-selection code I noticed a combination that seems worth flagging:
In Configuration.hs:
dmqcPeerSharing = I PeerSharingEnabled,
Peer sharing is enabled by default.
In PeerSelectionPolicy.hs:
-- | Trivial peer selection policy used as dummy value
--
policy :: ...
policy rngVar peerMetrics =
PeerSelectionPolicy {
...
policyMaxInProgressPeerShareReqs = 0,
policyPeerShareRetryTime = 0, -- seconds
policyPeerShareBatchWaitTime = 0, -- seconds
policyPeerShareOverallTimeout = 0, -- seconds
policyPeerShareActivationDelay = 2, -- seconds
...
}
The policy is explicitly commented as a "dummy value", and four of the peer-sharing-related timing/concurrency fields are zeroed out.
I also noticed that upstream ouroboros-network's own changelog states peer sharing "should be disabled by default" as a design principle, which doesn't match the default here.
Question: Is peer sharing intended to be functionally active in dmq-node today, governed by this dummy policy? Or is the PeerSharingEnabled default a placeholder that should be flipped to PeerSharingDisabled until the policy is properly tuned (similar to how the other promotion/demotion policies in this file have real logic)?
Happy to dig further or help test the actual runtime behavior if useful — just wanted to flag the inconsistency before assuming either direction.
While auditing the
diffusion/peer-selection codeI noticed a combination that seems worth flagging:In Configuration.hs:
In
PeerSelectionPolicy.hs:The policy is explicitly commented as a "dummy value", and four of the peer-sharing-related timing/concurrency fields are zeroed out.
I also noticed that upstream ouroboros-network's own changelog states peer sharing "should be disabled by default" as a design principle, which doesn't match the default here.
Question: Is peer sharing intended to be functionally active in dmq-node today, governed by this dummy policy? Or is the PeerSharingEnabled default a placeholder that should be flipped to PeerSharingDisabled until the policy is properly tuned (similar to how the other promotion/demotion policies in this file have real logic)?
Happy to dig further or help test the actual runtime behavior if useful — just wanted to flag the inconsistency before assuming either direction.