the DA bandwith of bitcoin is 4MB / 10 mins.
so it can handle 4 MBs every 10 minutes.
we right now keep adding transactions without ever checking our DA usage. on the sequencer side we only use the ongoing diff size for to make sequencer commitments.

Ideally, we'd like to have a fee mechanism that increases L1 fee rates on soft confirmations as we reach the 4MB/10 mins mark. The L1 fee is an arbitrary fee rate selected by the sequencer and only the sequencer. We have no rules at the moment regarding the verification of this fee rate.

So the sequencer can increase the fee rate independently of the fee rate on bitcoin arbitrarily.

A simple solution to this would be for example,

let mut l1_fee_rate = get_l1_fee_rate()

l1_fee_rate *= current_cumulative_diff_da_usage / BITCOIN_MAX_DA_BANDWITH

thought this is not really smart at all. :)

the idea here is that, as we reach the limit, we economically discourage transactions from happening.

Not having a maxL1FeeRate field on our transactions along with this update would mean some users would pay unexpected amounts of fees, however, we can look for a solution for this later on.

Do we want to consider a 4MB/10mins or have a moving target depending on current mempool usage ?

the idea here is that, as we reach the limit, we economically discourage transactions from happening.
Understand this but this will affect the TX distribution across blocks where we have our first l2 blocks filled and progressively empty.

What is the end goal of fee ?
Do we want to disincentivize TXs as we get closer to the limit ? Do we want to distribute the TXs across our blocks ? Fairness/predictability for users ?

What should happen if we hit the limit but don't make it to the l1 block in time of high congestion, as in early 2024 ?

We need to be mindful of re-orgs that might throw off calculations as well.

Do we want to consider a 4MB/10mins or have a moving target depending on current mempool usage ?

4MB/10mins.

What is the end goal of fee ?
Do we want to disincentivize TXs as we get closer to the limit ?

Yes. But the later the disincentivization starts, the better. Don't know how to set it up though :)

What should happen if we hit the limit but don't make it to the l1 block in time of high congestion, as in early 2024 ?

Things will get delayed. There's nothing we can do unfortunately. We can leave this as a later issue.