There is a dual model. If a group plans for full NOSTR interoperability then the longer signature is used, both are accepted on XPRS either way.
Then the choice is made once per group rather than once per packet, and that is the right place for it. A gateway that re-signs on the way into Nostr would be attesting to itself, not to the station: the packet would arrive perfectly verifiable and mean nothing about who sent it. Pushing the decision down to the group avoids that trap entirely. A group either promises interoperability and pays for it on every packet, or it does not and stays honest about being unverifiable from outside.
The arithmetic on the long format is worth stating plainly. Base85 turns four bytes into five characters, so a full BIP-340 signature costs 80 characters and the event id another 40. If the pubkey has to travel too, that is 160 of your 250 gone before a single tag or word, leaving 90 characters for everything a human would read. Keep the pubkey out because the group knows its members, and you hold on to 130. Not an argument against it, but it explains why the short format exists at all.
What I still cannot tell from the outside: is the format a flag on each packet, or a property of the group fixed at join time? If a station hears both kinds on the same channel, a neighbour running full signatures while the group runs short, does the receiver show the unverifiable one as non verified and carry on, or is the packet dropped as malformed? The second is the safer failure and the less friendly one.
Either way, deciding it at group level and accepting both on receipt is a cleaner answer than the one I expected. My question assumed a single format with a bridge left to improvise.
The arithmetic on the long format is worth stating plainly. Base85 turns four bytes into five characters, so a full BIP-340 signature costs 80 characters and the event id another 40. If the pubkey has to travel too, that is 160 of your 250 gone before a single tag or word, leaving 90 characters for everything a human would read. Keep the pubkey out because the group knows its members, and you hold on to 130. Not an argument against it, but it explains why the short format exists at all.
What I still cannot tell from the outside: is the format a flag on each packet, or a property of the group fixed at join time? If a station hears both kinds on the same channel, a neighbour running full signatures while the group runs short, does the receiver show the unverifiable one as non verified and carry on, or is the packet dropped as malformed? The second is the safer failure and the less friendly one.
Either way, deciding it at group level and accepting both on receipt is a cleaner answer than the one I expected. My question assumed a single format with a bridge left to improvise.
1