Tao Zhu

> @taozhu-chicago @t-nelson I was thinking we sort by fee, then bucket by account/CU limit. then forward by bucket. so if there are N buckets, we round robin between each...

> > Is it OK to use `FeeStructure::default()` instead of `bank.fee_structure` for fee calculation? > > FeeStructure is not fixed and will be modified at epoch boundaries over time. It...

Yea, so just cache `fee_structure` and `lamports_per_signature`

the bucket is by writable-account. In your example above, after sorting, we'd have a view of buffer like this: [(A, 1000000), (AB, 1000000), (A, 1), (B, 1), (A, 1), (B,...

> > the bucket is by writable-account. > > Hmmm, does this mean every account that tries to grab a write lock on `A` will fall into the same bucket?...

1. Bucket by write lock is to gate TXs not to exceed max_account_limit. (A minor correction is `AB` will be in `bucket 2` as well, as it will also take...

Hmm, isn't this why we have block-limit and account-limit to help such scenario? Plus allowing additional-fee to give `AB` better chance? I feel I am missing something. Can you describe...

Like the idea to use same/similar market mechanism to handle fee (cost of computing) and account size (cost of storage). It is doable from implementation perspective. My intuition is safer...

@CriesofCarrots I don't have actual RPC query use cases, I imagine user would construct a transaction, and know a list of write accounts. Then user would ask something like "what's...

> One note: I can't decide whether to pluralize "fees" when referring to the collection struct. There might only be one value (what you call the block_fee, but I'd like...