I was checking DUSK gas settings and stopped at one number: 8 million. The default hook gets 250,000 gas. That’s a 32× range; at 10 gwei, 0.08 Ethereum versus 0.0025 ETH.
A Covenant’s condition occupies just 52 bytes, expressing price, time, or both. Elegant. Also restrictive. I’d test ten application types and count how many fit without moving complexity into hooks.
Then precision: exact USD strings allow eight decimal places. `2000.12345678` works; floating-point inputs or extra digits fail validation. A tiny rounding mistake could undermine an otherwise secure instruction.
This matters for Nillon because accepted callbacks aren't completed trades. Refunds can fail their 30,000-gas push and sit unclaimed until the payer acts.
Retry windows stretch from seven to thirty days. Fine, but who retries? Measure recovery after one hour, 24 hours, seven days and thirty days. Nodes also depend on archive-capable RPCs handling 10,000-block log queries. Resync across a million blocks isn't free.
Before trusting Nillon usage numbers, I’d measure hook outcomes, outstanding refunds, RPC failures, and CPU/RAM at 100, 500 and 1,000 Covenants.
With Nillon I keep coming back to the same gap: the instruction can stay private while execution quality remains painfully measurable.
#NIL $NIL @Nillion
A Covenant’s condition occupies just 52 bytes, expressing price, time, or both. Elegant. Also restrictive. I’d test ten application types and count how many fit without moving complexity into hooks.
Then precision: exact USD strings allow eight decimal places. `2000.12345678` works; floating-point inputs or extra digits fail validation. A tiny rounding mistake could undermine an otherwise secure instruction.
This matters for Nillon because accepted callbacks aren't completed trades. Refunds can fail their 30,000-gas push and sit unclaimed until the payer acts.
Retry windows stretch from seven to thirty days. Fine, but who retries? Measure recovery after one hour, 24 hours, seven days and thirty days. Nodes also depend on archive-capable RPCs handling 10,000-block log queries. Resync across a million blocks isn't free.
Before trusting Nillon usage numbers, I’d measure hook outcomes, outstanding refunds, RPC failures, and CPU/RAM at 100, 500 and 1,000 Covenants.
With Nillon I keep coming back to the same gap: the instruction can stay private while execution quality remains painfully measurable.
#NIL $NIL @Nillion