<p​lowsof:matrix.org> some feedback to share (using gist as a pastebin): 1. initial feedback from CypherStack for the latest bp++ paper gist.github.com/plowsof/534778636eca474951e4661507cdc205 . 2. Rough scope of work for the seraphis papers which j-berman / UkoeHB / Kayabanerve helped create gist.github.com/plowsof/8cb33e2efe4bf0239927ad3bd92326e0 ( which if no one has issues with, ca<clipped message>

<p​lowsof:matrix.org> n be sent out for quotes )

Do we currently use batch verification?

Yes, but not as much as we could.

"While the preprint mentions that BP++ supports batch verification, it provides no details on the corresponding algebra" <-- this will need clarification, because otherwise I imagine BP++ without batch verification might end up slower?

Batch verification should be pretty easy to implement, since it’s just about sharing generators between proofs.

AFAIK, Cypherstack's job is supposed to check the validity of the arguments that are presented in the paper. Not to create anything new that's not in the paper. So it wouldn't be their job to figure out batch verification for BP++.

When can batch verification be used? WIt would seem easy when syncing blocks. WHat about transaction relay? Would the node wait for "enough" candidate txs to arrive and then check them in a batch? That would slow down tx propagation.

Transactions that use the Curve Trees idea would be slow to verify if done individually, according to the data on verification time in the paper.

They are batch-verified at the block level currently