references:
In the Mantle protocol, a Channel is a virtual chain maintained by one or more sequencers on behalf of a Zone. Each channel maintains a state that includes a list of accredited keys — the set of sequencers authorized to write messages to it.
The Round-Robin algorithm is the mechanism Mantle uses to determine, at any
given block slot, which sequencer in the accredited list holds the current write
authority. Only the sequencer holding write authority at a given slot is permitted
to post a valid CHANNEL_INSCRIBE operation to that channel.
The full implementation from the Mantle specification is reproduced below:
# Round Robin algorithm determining the new sequencer index and the
# new sequencer starting slot
def round_robin(block_slot: Slot, channel: ChannelState) -> (u16, u64):
elapsed_slots = block_slot - channel.tip_slot
if (elapsed_slots >= channel.posting_timeout && channel.posting_timeout != 0):
# Get the number of sequencers that get timed out
sequencers_timed_out = elapsed_slots // channel.posting_timeout
index = (channel.tip_sequencer + sequencers_timed_out) \
% len(channel.accredited_keys)
starting_slot = channel.tip_slot \
+ sequencers_timed_out * channel.posting_timeout
elif channel.posting_timeframe != 0:
# Get the number of timeframes elapsed to get who is the sequencer
tip_sequencer_duration = block_slot - channel.tip_sequencer_starting_slot
rotations = tip_sequencer_duration // channel.posting_timeframe
index = (channel.tip_sequencer + rotations) \
% len(channel.accredited_keys)
starting_slot = channel.tip_sequencer_starting_slot \
+ rotations * channel.posting_timeframe
else:
# Infinite turn (posting_timeframe = 0), not timed out
index = channel.tip_sequencer
starting_slot = channel.tip_sequencer_starting_slot
return (index, starting_slot)
This mode is activated when:
posting_timeout > 0, ANDelapsed_slots >= posting_timeoutThat is, the last message was posted far enough in the past that the current sequencer is considered to have exceeded its allowed idle period.
When a sequencer fails to post a message within posting_timeout slots since
the last message was written, the algorithm treats this as an implicit forfeit.
It then computes how many consecutive sequencers have been skipped due to
inactivity:
sequencers_timed_out = elapsed_slots // channel.posting_timeout
index = (channel.tip_sequencer + sequencers_timed_out)
% len(channel.accredited_keys)