Invalid transaction against best block vs finalized block - status propagation

[jsamol]

Hi! I've noticed that Smoldot clears invalid transactions and reports back on them only when the block they've been verified against gets promoted from the best to the finalized block. This results in a noticeable delay between submission and arrival of the rejected status update compared to what I observe when I'm communicating directly with a Substrate node. Additionally, I've observed that on some less stable chains the transaction rejection sometimes never gets propagated to the point that is reported back.

I briefly looked at what the Substrate node does and, if I haven't missed anything, it appears that it returns the error immediately after validating against the best block.

Provided that I'm correct with my code interpretation, where does this difference between how Smoldot and Substrate handle transaction verification come from?

[josepot]

Hi! I've noticed that Smoldot clears invalid transactions and reports back on them only when the block they've been verified against gets promoted from the best to the finalized block.

This is the correct behaviour, though. The transaction could have been built on top of a different block that may later become the best/finalized one.

Please keep in mind that the consumer can also handle the validation, the broadcasting and the tracking of the transaction... That's actually what PAPI does, we only use transaction_v1_broadcast and then we handle the validation and the tracking ourselves. That also allows us to provide all the feedback that the user needs ASAP.