With stateless client, the cost of reshuffling would be reduced massively. It gives us a chance to remove shuffling cycle.
The other problem is making the whole procedure fast enough to collate the collation.
So once the validator finds he was sampled as the collator of this period:
Check if he's watching this shard
if not, do the fast syncing and ask for transactions of this shard
Collate
Send add_header tx and broadcast the collation
Since we set the period length to 5 blocks, so the add_header tx must be included in Mainnet in 75 seconds.
With stateless client, the cost of reshuffling would be reduced massively. It gives us a chance to remove shuffling cycle.
The other problem is making the whole procedure fast enough to collate the collation. So once the validator finds he was sampled as the collator of this period:
add_header
tx and broadcast the collationSince we set the period length to 5 blocks, so the
add_header
tx must be included in Mainnet in 75 seconds.