[WIP] Minimal ePBS without Max EB and 7002 #2

potuz · 2024-04-02T18:50:18Z

Full specification for ePBS

The full design notes are included in https://hackmd.io/@potuz/rJ9GCnT1C
Forkchoice annotated spec can be found in https://hackmd.io/@potuz/SJdXM43x0
Currently fully specified documents:

beacon-chain.md
p2p-interface.md
validator.md
builder.md
fork-choice.md
engine.md

Most of the engine API and the forkchoice handlers in forkchoice.md are ready

Still missing is:

Execution specification

Also notice that the DA issue only happens if all the transactions are invalidated in the same slot

Notice that the whole thing is broken by the construct if index < len(validators) else.

- Added handlers for execution payloads and checks inclusion lists availability on consensus blocks

Move the builder id and other unnecessary information outside of the payload.

come-maiz

Thanks for all this work and the care you have put on branch feature.
I've left a few comments during my initial reading.

specs/_features/epbs/beacon-chain.md

come-maiz · 2024-04-22T12:28:23Z

specs/_features/epbs/beacon-chain.md

+At any given slot, the status of the blockchain's head may be either 
+- A block from a previous slot (eg. the current slot's proposer did not submit its block). 
+- An *empty* block from the current slot (eg. the proposer submitted a timely block, but the builder did not reveal the payload on time). 
+- A full block for the current slot (both the proposer and the builder revealed on time). 


Suggested change

- A full block for the current slot (both the proposer and the builder revealed on time).

- A block with payload for the current slot (both the proposer and the builder revealed on time).

Full can be confusing. Is the block full of transactions? Instead of empty or full, what about block with or without payload?

Full is used throughout the whole list of long documents I worked on, changing this will be overkill for this stage of the project

come-maiz · 2024-04-22T12:29:42Z

specs/_features/epbs/beacon-chain.md

+| `PAYLOAD_ABSENT` | `uint8(0)` |
+| `PAYLOAD_PRESENT` | `uint8(1)` | 
+| `PAYLOAD_WITHHELD` | `uint8(2)` | 
+| `PAYLOAD_INVALID_STATUS` | `uint8(3)` |


Thanks for all the reading, I'll go over them and change the typos. I won't touch anything about naming cause there's always a lot of bikeshedding about that. I'll defer naming discussions until the very last moment when I am no longer involved

specs/_features/epbs/validator.md

Left out naming suggestions until a later stage. Co-authored-by: Leonardo Arias <[email protected]> Co-authored-by: Ryan Schneider <[email protected]>

Ruteri · 2024-04-27T08:30:57Z

Thanks for all the effort you are putting into solving this nasty problem!

I think I'm missing some context on the payload attestation committee (PTC) after reading both the annotated spec and spec notes. Could you point me to where incentives for the committee are discussed? What incentivises PTC to vote truthfully and timely - and vote in the first place?
Related - what is the probability colluding proposers (say 30% of the stake) control the majority of the PTC votes for a slot? This is important since a majority PTC seems to be able to coerce either the proposer to include a particular payload or the builders to pay them by otherwise voting untruthfully.

It seems like inclusion lists could be used to force mayhem in periods of high MEV. Could you help me see the reason why the proposer for slot N (that possibly also controls slot N+2 or colludes with someone who does which happens) would not force the proposer in slot N+1 to include 1 wei priority gas fee 30 million gas transaction? This would mean the proposer controlling both N and N+2 can extract all the value available in slots N, N+1 and N+2 which would be disastrous.

potuz · 2024-04-29T10:55:31Z

Could you point me to where incentives for the committee are discussed? What incentivises PTC to vote truthfully and timely - and vote in the first place?

PTC attesters get rewarded a full attestation reward if they attest correctly and get penalized as a missed attestation if they don't.

Related - what is the probability colluding proposers (say 30% of the stake) control the majority of the PTC votes for a slot?

In order to save time let me do a larger case, in which the attacker has 35% of the stake (so that python's default float does not overflow), in this case there's a probability of ~10^-12 per slot:

>>> from scipy.stats import binom
>>> 1 - binom.cdf(k=255, n=512, p=0.35)
2.461586490198897e-12

Which gives an expected time of about 154 582 years to succeed.

... would not force the proposer in slot N+1 to include 1 wei priority gas fee 30 million gas transaction?

The maximum gas of the IL is 3M and this is on top of any gas limit for the next block.

Ruteri · 2024-05-13T15:12:06Z

Thanks for all the answers!

PTC attesters get rewarded a full attestation reward if they attest correctly and get penalized as a missed attestation if they don't.

Follow up: since attesters are penalized for missing votes, wouldn't they be incentivized to vote "missing payload" as soon as they can to guarantee they don't miss the vote? Is the reward for voting the same in all cases?

The maximum gas of the IL is 3M and this is on top of any gas limit for the next block.

Ah I see. I wonder if there's anything funny one can do with nonces - say there are two transactions with the same nonce. One can try and time releasing the transactions such that they propagate to some fraction of PTC

potuz · 2024-05-15T15:37:17Z

Follow up: since attesters are penalized for missing votes, wouldn't they be incentivized to vote "missing payload" as soon as they can to guarantee they don't miss the vote? Is the reward for voting the same in all cases?

If the PTC member attest for "missing payload" but the majority is honest, the payload is present and becomes canonical, the PTC member will be penalized. The reward is only applied when voting for the right outcome.

g11tech · 2024-06-21T09:45:53Z

specs/_features/epbs/beacon-chain.md

+    state_root: Bytes32
+    receipts_root: Bytes32
+    logs_bloom: ByteVector[BYTES_PER_LOGS_BLOOM]
+    prev_randao: Bytes32  # 'difficulty' in the yellow paper


Suggested change

prev_randao: Bytes32 # 'difficulty' in the yellow paper

prev_randao: Bytes32 # 'mix_hash' in the yellow paper

has this changed in the YP and the main repo? I took this one out from the main repo

g11tech · 2024-06-21T09:51:51Z

specs/_features/epbs/beacon-chain.md

+
+#### `ExecutionPayloadHeader`
+
+**Note:** The `ExecutionPayloadHeader` is modified to only contain the block hash of the committed `ExecutionPayload` in addition to the builder's payment information and KZG commitments root to verify the inclusion proofs. 


one consideration of compression here by dropping other fields are if there are any apps/use cases are building/consuming proofs against the other attributes of payload against the inclusion in a block

One way to make this easy would be to have a merkle hash of the execution payload ad the block hash here

(just thinking aloud)

g11tech · 2024-06-21T10:42:25Z

specs/_features/epbs/builder.md

+
+Builders can broadcast a payload bid for the current or the next slot's proposer to include. They produce a `SignedExecutionPayloadHeader` as follows. 
+
+Prior to constructing a payload, the builder **MUST** have a full `InclusionList` object from the proposer matching `state.previous_inclusion_list_proposer`. The signed summary for this inclusion list will be needed when revealing the full execution payload (see below). 


may be totally remove inclusion list for now?

it will be removed in the EIP, I'll open a new PR in this repo based on Electra soon

* Add fork logic

Ali1994Baker · 2024-08-11T05:47:04Z

Ali1994baker

Ali1994Baker · 2024-08-11T06:00:06Z

Ali1994baker

potuz force-pushed the epbs_stripped_out branch 4 times, most recently from 46432e7 to ac34ffc Compare April 3, 2024 14:10

potuz added 26 commits April 3, 2024 14:28

init

5ef39b8

Add engine methods and check IL validity

1a49a51

add small design notes for ILs

0e2267d

check for gas usage after payload insertion

844381d

Also notice that the DA issue only happens if all the transactions are invalidated in the same slot

Add helpers that change on ePBS

ee37d4c

switch to gas limit instead of usage as pointed by V. Buterin

13d75bc

Added more changed helpers

f4b566b

Notice that the whole thing is broken by the construct if index < len(validators) else.

change design notes

08b881d

minor changes

d339dfc

First forkchoice changes.

ff1eff7

- Added handlers for execution payloads and checks inclusion lists availability on consensus blocks

don't modify verify_block_signature

ef7bd40

remove clutter

9dbd2c5

Check IL compatibility with parent block hash

c7331a1

Do not broadcast the parent_block_hash but use it as parameter to the EL

8d6ce8b

process payment immediately

759cdd6

Deal with withdrawals

90b42e6

fix withdrawals, check only in the CL

a604462

remove unused request dataclass

d6ca928

pass the propsoer_index to process_execution_payload_header

70a6e6d

run doctoc

2c4fc84

only cache the withdrawals root in the state

43a8dd6

ptc rewards pass 1

52067d5

only pass relevant IL info to the EL

a3c2468

Use envelopes instead of adding builder info to the payload

800a910

Move the builder id and other unnecessary information outside of the payload.

ptc message handlers in forkchoice

0228ea0

take ptc members from all committees

1ae9cd9

potuz and others added 16 commits April 17, 2024 11:44

consider PTC vote untying max weight

8674517

Fix boost computation

b2c6482

use ChildNode across forkchoice

17b9e90

doctoc

c63b81e

modify get_ancestor to return a ChildNode

5ff8d54

typo

f35cc68

fix boosts for future slots

18c31e4

remove unnecessary node redefinition

ba0283c

descend only into right payload except at the justified checkpoint

9edc282

update head root in get_head loop

f3a035a

update get_head to use ChildNodes

6c60686

stranded on_block simplifications

502e98d

did not remove process_withdrawals

05fbb9a

Add engine_getInclusionListV1 to engine.md

1e9406d

coherent naming in engine API

8883b29

fix unused constant

ae35416

come-maiz reviewed Apr 22, 2024

View reviewed changes

potuz and others added 2 commits April 24, 2024 13:32

Apply suggestions from Leonardo Arias

eedf7fb

Left out naming suggestions until a later stage. Co-authored-by: Leonardo Arias <[email protected]> Co-authored-by: Ryan Schneider <[email protected]>

Move exec payload header position to replace latest payload header

5ebda89

require the target blockroot of a PTC message is valid

2decd43

g11tech reviewed Jun 21, 2024

View reviewed changes

Add fork logic for ePBS (#6)

d0060c0

* Add fork logic

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] Minimal ePBS without Max EB and 7002 #2

[WIP] Minimal ePBS without Max EB and 7002 #2

potuz commented Apr 2, 2024 •

edited

Loading

come-maiz left a comment

come-maiz Apr 22, 2024

potuz Apr 24, 2024

come-maiz Apr 22, 2024

potuz Apr 24, 2024

Ruteri commented Apr 27, 2024

potuz commented Apr 29, 2024

Ruteri commented May 13, 2024

potuz commented May 15, 2024

g11tech Jun 21, 2024

potuz Jun 24, 2024

g11tech Jun 21, 2024

g11tech Jun 21, 2024

potuz Jun 24, 2024

Ali1994Baker commented Aug 11, 2024

Ali1994Baker commented Aug 11, 2024

	- A full block for the current slot (both the proposer and the builder revealed on time).
	- A block with payload for the current slot (both the proposer and the builder revealed on time).

	\| `PAYLOAD_INVALID_STATUS` \| `uint8(3)` \|
	\| `PAYLOAD_INVALID` \| `uint8(3)` \|

	prev_randao: Bytes32 # 'difficulty' in the yellow paper
	prev_randao: Bytes32 # 'mix_hash' in the yellow paper


		#### `ExecutionPayloadHeader`

		Note: The `ExecutionPayloadHeader` is modified to only contain the block hash of the committed `ExecutionPayload` in addition to the builder's payment information and KZG commitments root to verify the inclusion proofs.


		Builders can broadcast a payload bid for the current or the next slot's proposer to include. They produce a `SignedExecutionPayloadHeader` as follows.

		Prior to constructing a payload, the builder MUST have a full `InclusionList` object from the proposer matching `state.previous_inclusion_list_proposer`. The signed summary for this inclusion list will be needed when revealing the full execution payload (see below).

[WIP] Minimal ePBS without Max EB and 7002 #2

Are you sure you want to change the base?

[WIP] Minimal ePBS without Max EB and 7002 #2

Conversation

potuz commented Apr 2, 2024 • edited Loading

come-maiz left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Ruteri commented Apr 27, 2024

potuz commented Apr 29, 2024

Ruteri commented May 13, 2024

potuz commented May 15, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Ali1994Baker commented Aug 11, 2024

Ali1994Baker commented Aug 11, 2024

potuz commented Apr 2, 2024 •

edited

Loading