Currently, Celestia’s minimum gas price is set to a level where DA fees are basically nothing, which is currently enforced by the default minimum mempool fee. A consensus-critical min fee can be enforced using CIP-6.
Fees currently today are dirt cheap (basically ~free), as the current goal of fees is to prevent spam, rather than generate protocol revenue. DA fees for a typical rollup transaction today is around $0.00002, with Celestia charging around $0.08 per MB, around 1000x cheaper than EIP-4844.
The purpose of this thread is to start a discussion about DA fees price discovery and when or if we should increase DA fees.
The table below shows the impact of increasing fees, assuming a TIA price of $4.
Multiplier
Price per MB (TIA)
Price per MB (USD)
DA fee per 256 byte rollup tx (USD)
1x (today)
0.02
0.08
0.00002048
2x
0.04
0.16
0.00004096
4x
0.08
0.32
0.00008192
8x
0.16
0.64
0.00016384
16x
0.32
1.28
0.00032768
32x
0.64
2.56
0.00065536
64x
1.28
5.12
0.00131072
128x
2.56
10.24
0.00262144
I believe a short-term increase of up to ~4x from today’s minimum fee could allow the protocol to test the waters in DA fee price discovery, while still keeping DA fees extremely low at around $0.00008 per rollup transaction, and fees would still be very low even if TIA price increased.
what do you think you can learn from testing the waters at this stage? it’s hard for me to imagine any of the current crop of rollups leaving because of a 4x increase. do you feel there’s a reasonable chance that’s not true?
i do however worry that it signals a philosophical shift that might make those who are currently on the fence more wary of committing to celestia. so there could be a path dependence here (there might not be one but it’s at least worth thinking about)
i would personally lean towards not touching fees until you feel you have sufficiently strong network effects (i.e a sufficiently distributed and loyal community – which probably means several rollups of eclipse’s size + (sub)communities that really live and breathe tia) – unless there’s a clear pressing reason to do so beforehand
weakly held opinion though, and it’s probable i’m missing something
There is probably room to increase DA fees by up to 100x ($0.001) per transaction, while DA still being only 1% of the cost of a reasonable $0.01 end-user transaction fee once you add in compute costs etc. https://x.com/musalbas/status/1893975544636563736
100x increase does keep Celestia rollups competitive with scaleable L1s, however don’t we wish to enable use cases which aren’t feasible with today’s fees?
$0.01 is reasonable for a financial transaction, but not for a a social rollapp where every like tips $0.01, the right fee for that is <$0.001
For 4x
Rollups are incentivized to appreciate DA efficiency (such as compression).
Cheap enough to not hurt induced demand, even if TIA quadruples.
Road to profitability becomes less delusional.
Risk
Community might get distracted and develop an appetite for more fees, while we’re still so early in the scaling journey.
During 2021-2023, many in the Ethereum community were distracted admiring the burn instead of panicking about not having scaled yet.
Personally I don’t think this will happen to Celestia and do believe Ethereum will make it.
One issue with analyzing blockspace economics of Celestia is that blockspace is priced in TIA which is very volatile. It’s likely that Celestia’s users will want to think about and project blockspace cost in USD terms.
In fact, an increase of 4x as proposed here would be achieved simply by the price of TIA increasing by 4x. Then if we had also increased the base fees by 4x, the total increase would be 16x which could possibly push cost-sensitive rollups off of Celestia. Or in the opposite scenario, if the price of TIA decreased by 4x then the protocol revenue would also decrease by 4x. This doesn’t seem like a very stable economic model to build on top off.
For this reason I think that before we talk about how much the minimum fee for blockspace should be, we consider pricing Celestia blockspace in USD rather than TIA.
To be clear, I mean that the fees should still be payable in TIA not USD, but the amount of TIA per MB should dynamically target a fixed USD price.
minimum and maximum price as well. minimum will support the protocol and max the consumers. knowing the pricing threshold for a fixed period of time will certainly help.
Good question. The approaches that have been thrown around for this include enshrining an oracle (like you mention), reading the price from some TIA/USD pool, or letting the validators vote periodically to adjust the min fee (like you mention). I don’t think there’s a clear answer yet nor is this particularly urgent.
This whole idea could be moot though if in the future Celestia blockspace is congested since then we’ll have to rely on some other pricing mechanism. I think in the success case for Celestia some degree of congestion is highly likely.
In that case, if we want to provide predictable pricing for Celestia users we might need to have blockspace futures.
As a proof of concept I think it would be useful for Celestia to test this fee increase. One one hand it could actually prove to the rollups that the fee is negligible even when the fee is increased. On the other hand it does signal that Celestia is serious about monetizing the blockspace. I looked back at the pull request and it seems like the min fee was chosen somewhat arbitrary. Finally I do think that frequent changes would be perceived as bad and I do agree with Nick that the fee should be set in USD.
`
[quote="sacha, post:2, topic:1924"]
i do however worry that it signals a philosophical shift that might make those who are currently on the fence more wary of committing to celestia. so there could be a path dependence here (there might not be one but it’s at least worth thinking about)
[/quote]
`
I am curious what kind of feedback you all are getting when protocols are on the fence about choosing a DA layer?
