Overall, LGTM. I requested we discuss this on ACD 151.

What is the use case for this? These days I'm hesitant to proactively add things that don't have incredibly compelling use cases because once added, changing/removing become hard. When we switch to verkle trees or make other changes the account structure may change so I'm particularly hesitant to expose all of this as it feels like a leaky abstraction.

If the intent is for debugging purposes, can we add it to a debug_ namespace rather than eth_ namespace, to make it clear that dapps shouldn't rely on the content being consistent long term?

If the intent is for debugging purposes, can we add it to a debug_ namespace rather than eth_ namespace

I'd be fine with that

I would prefer to see storageRoot rather than just root.

Also, I see this as useful for more than just debugging, so I'm ok with the eth_ namespace.

Also, I see this as useful for more than just debugging, so I'm ok with the eth_ namespace.

Do you have some example use cases other than debugging?

Also, I see this as useful for more than just debugging, so I'm ok with the eth_ namespace.

Do you have some example use cases other than debugging?

Optimism commits to the storage root of a particular L2 contract on L1 for withdrawals. A user must provide a storage proof to finalize the withdrawal. We currently get the storage root with eth_getProof.

Also, I see this as useful for more than just debugging,

I can live with either, but I do think this is information should be a first-class citizen of eth. It's a very basic unit of information, and can be useful for a variety of purposes.

We currently get the storage root with eth_getProof.

However, getting things using eth_getProof sounds like a safer solution though

I would prefer to see storageRoot rather than just root.

Makes sense to me

Optimism commits to the storage root of a particular L2 contract on L1 for withdrawals. A user must provide a storage proof to finalize the withdrawal. We currently get the storage root with eth_getProof.

What problem are you having with eth_getProof that would be solved by eth_account?

What problem are you having with eth_getProof

eth_getProof solves a different class of problems. It also requires a lot more 'dedication' from the server, which needs to traverse the trie and look up proofs to serve the request. It is the same role a legacy light server serves, and it cannot be served by a node which is trie-less, does aggressive pruning or generally has a fast direct lookup and slow trie-lookup.

SO... Geth (as represented by me and @lightclient ) is on board behind this, so is Besu (via @shemnon ) -- what do we need to do to move forward?

I'd like to get a 👍 from nethermind and erigon. I requested we mention it on the next ACD. If they thumb up on this PR out-of-band I think we can just merge.

Erigon doesn't store historical MPT roots, so we can return storageRoot only for the current/latest state.

the question has just been asked on ACD: verkle trees won't be able to return storageRoot as this will no longer exist.

Seems like 2 main pieces of feedback from ACD:

• many clients doesn't have storageRoot very far behind head, including erigon, so I'm not sure how useful that value will be
• how will this play with verkle trees?

many clients doesn't have storageRoot very far behind head, including erigon, so I'm not sure how useful that value will be

This does not really compute for me. Please elaborate. Do you mean that in order to produce the consensus-definition of an account, they have to iterate the (flat) storage? If so, I'm curious which clients, apart from Erigon. I really do not think (hope) this is a common way to handle account storage.

This does not really compute for me. Please elaborate. Do you mean that in order to produce the consensus-definition of an account, they have to iterate the (flat) storage? If so, I'm curious which clients, apart from Erigon. I really do not think (hope) this is a common way to handle account storage

On that note, any thoughts now or when Verkle to retrieve the whole account storage using the same method?

, any thoughts now or when Verkle to retrieve the whole account storage using the same method?

Well, leaving verkle aside for now, retrieving the whole storage is not "generally" doable. For clients which have only a trie-based represenation, returning the whole storage might mean iterating over a trie containing hundreds of thousands of leaves. For clients with a flat storage representation, it is "simpler" but still, for some accounts, the storage portions are huge. And returning those over RPC is a huge job. Anyway, it's not really related to this feature

Well, leaving verkle aside for now, retrieving the whole storage is not "generally" doable. For clients which have only a trie-based represenation, returning the whole storage might mean iterating over a trie containing hundreds of thousands of leaves. For clients with a flat storage representation, it is "simpler" but still, for some accounts, the storage portions are huge. And returning those over RPC is a huge job. Anyway, it's not really related to this feature

@holiman Thank you, yes you are right if you store only the trie representation it is pretty impossible, hence the Verkle comment as potential improvement / thought / hope :). But you are right some contracts will have a huge state too, so not great for an rpc call. My thought was for users not needing to constantly request data if they could verify it has not changed locally (cached), but bringing down the whole state will negate the benefit on big accounts / contracts as opposed to retrieving a specific value using account_proofs, and yes not related to this either.

Coming back to this, I think given the fact that i) all this information is available via other means, such as eth_getProof and ii) verkle does not have the concept of storageRoot and we would need update this method for that, I am inclined to simply close this.

I guess last call- is there a reason to continue pursuing this versus using the existing methods given the tradeoffs?

Fair enough.. Closing