Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Extrinsic v5 definition and specification #124

Open
wants to merge 6 commits into
base: main
Choose a base branch
from
Open
Changes from 4 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
155 changes: 155 additions & 0 deletions text/0124-extrinsic-version-5.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,155 @@
# RFC-0124: Extrinsic version 5

| | |
| --------------- | ------------------------------------------------------------------------------------------- |
| **Start Date** | 18 October 2024 |
| **Description** | Definition and specification of version 5 extrinsics |
| **Authors** | George Pisaltu |

## Summary

This RFC proposes the definition of version 5 extrinsics along with changes to the specification and encoding from version 4.

## Motivation

[RFC84](https://github.com/polkadot-fellows/RFCs/blob/main/text/0084-general-transaction-extrinsic-format.md) introduced the specification of `General` transactions, a new type of extrinsic besides the `Signed` and `Unsigned` variants available previously in version 4. Additionally, [RFC99](https://github.com/polkadot-fellows/RFCs/blob/main/text/0099-transaction-extension-version.md) introduced versioning of transaction extensions through an extra byte in the extrinsic encoding. Both of these changes require an extrinsic format version bump as both the semantics around extensions as well as the actual encoding of extrinsics need to change to accommodate these new features.

## Stakeholders

- Runtime users
- Runtime devs
- Wallet devs

## Explanation

### Changes to extrinsic authorization

The introduction of `General` transactions allows the authorization of any and all origins through
extensions. This means that, with the appropriate extension, `General` transactions are capable of
replicating the same behavior present day v4 `Signed` transactions. Specifically for Polkadot
chains, an example implementation for such an extension is
[`VerifySignature`](https://github.com/paritytech/polkadot-sdk/tree/master/substrate/frame/verify-signature),
introduced in the Transaction Extension
[PR3685](https://github.com/paritytech/polkadot-sdk/pull/3685). Other extensions can be inserted
into the extension pipeline to authorize different custom origins. Therefore, a `Signed` extrinsic
variant is redundant to a `General` one strictly in terms of functionality available to users and
would eventually need to be deprecated and removed.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Besides some wording suggestions, I don't think we should introduce links to polkadot-sdk repo. The idea of authorization through transaction extensions is already clear without the example IMO.

Suggested change
The introduction of `General` transactions allows the authorization of any and all origins through
extensions. This means that, with the appropriate extension, `General` transactions are capable of
replicating the same behavior present day v4 `Signed` transactions. Specifically for Polkadot
chains, an example implementation for such an extension is
[`VerifySignature`](https://github.com/paritytech/polkadot-sdk/tree/master/substrate/frame/verify-signature),
introduced in the Transaction Extension
[PR3685](https://github.com/paritytech/polkadot-sdk/pull/3685). Other extensions can be inserted
into the extension pipeline to authorize different custom origins. Therefore, a `Signed` extrinsic
variant is redundant to a `General` one strictly in terms of functionality available to users and
would eventually need to be deprecated and removed.
The introduction of `General` transactions allows the authorization of any origin through transaction extensions. This means that, with the appropriate extension, `General` transactions can replicate the same behaviour in present-day v4 `Signed` transactions. Therefore, a `Signed` extrinsic variant is redundant to a `General` one strictly in terms of user functionality and could eventually be deprecated and removed.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is in the explanation section and I think it adds valuable context to the discussion. I appreciate that you're familiar with the idea of authorization through transaction extensions, but I think it's still a novel topic and the example is useful. However, if others disagree, I will remove it.


### Encoding format for version 5

As with version 4, the encoded v5 extrinsics will still be an array of SCALE encoded bytes, starting
with the encoded length of the following bytes. The leading byte will determine the version and type
of extrinsic, as specified by
[RFC84](https://github.com/polkadot-fellows/RFCs/blob/main/text/0084-general-transaction-extrinsic-format.md),
with the addition that the `Signed` variant will not be supported for v5 extrinsics, for reasons
mentioned above.
georgepisaltu marked this conversation as resolved.
Show resolved Hide resolved

NOTE: For `Bare` extrinsics, the following bytes will just be the encoded call and nothing else.

For `General` transactions, as stated in
[RFC99](https://github.com/polkadot-fellows/RFCs/blob/main/text/0099-transaction-extension-version.md),
an extension version byte must be added in the next extrinsic version. This byte should allow
georgepisaltu marked this conversation as resolved.
Show resolved Hide resolved
runtimes to expose more than one set of extensions which can be used for a transaction. As far as
the v5 extrinsic encoding is concerned, this extension byte should be encoded immediately after the
leading encoding byte. The extension version byte should be included in payloads to be signed by all
extensions configured by runtime devs to ensure a user's extension version choice cannot be altered
by third parties.

After the extension version byte, the extensions will be encoded next, followed by the call itself.

A quick visualization of the encoding:

- `Bare` extrinsics: `(extrinsic_encoded_len, 0b0000_0101, call)`
- `General` transactions: `(extrinsic_encoded_len, , 0b0100_0101, extension_version_byte, extension, call)`
georgepisaltu marked this conversation as resolved.
Show resolved Hide resolved
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I thought the component is called extensions?

Suggested change
- `General` transactions: `(extrinsic_encoded_len, , 0b0100_0101, extension_version_byte, extension, call)`
- `General` transactions: `(extrinsic_encoded_len, , 0b0100_0101, extensions_version_byte, extensions, call)`

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The extension itself is usually a tuple of multiple extensions, generally referred to as the extension pipeline. Technically it's only one extension, the TxExtension commonly defined in runtimes, but that is always a tuple of extensions like CheckNonce, CheckWeight, ChargeTransactionPayment etc., so it would be only one extension version, as it is the version of the tuple, but there are multiple extensions in the pipeline.


### Signatures on Polkadot in General transactions

As stated before, [PR3685](https://github.com/paritytech/polkadot-sdk/pull/3685) comes with a
Transaction Extension which replicates the current `Signed` transactions in v5 extrinsics, namely
[`VerifySignature`](https://github.com/paritytech/polkadot-sdk/tree/master/substrate/frame/verify-signature).
This extension leverages the new inherited implication functionality introduced in
`TransactionExtension` and creates a payload to be signed using the data of all extensions after
itself in the extension pipeline. In order to run a transaction with a signed origin, a user must
create the transaction with an instance of the extension which provides a signature. Alternatively,
if users want to use some other origin, they should create the transaction with this particular
extension disabled. More on this behavior in the extension documentation. This extension can be
configured to accept a `MultiSignature`, which makes it compatible with all signature types
currently used in Polkadot.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same as before, this should not explain HOW polkadot-sdk will use the interface introduced in this RFC, but rather the shape of it and its goals. This paragraph adds very little value on that regard IMO.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The goal of it is to explain how to use the new v5 format to achieve provide the same functionality as signed transactions. I then use the example in polkadot-sdk to show this. The previous phrasing was a bit off and the intent wasn't clear, hopefully now it's better.

I don't think it's wrong to have links to polkadot-sdk or reference stuff from there. polkadot-sdk is used right now in Polkadot and Kusama runtimes, and quite extensively I might add. Even if we might not want to use it to define future functionality, it's inevitable that describing current functionality will point to polkadot-sdk. The upstream/consumer dependency is already there, I think making the RFC overly general by avoiding specific examples doesn't help in any way, it just makes it harder to read and understand. This does not in any way mean that the example pulled from polkadot-sdk is the only accepted implementation of specified functionality, it just helps me not unfurl a lot of code in this file.

That said, if others consider the examples are not useful, I will remove them.


To generate the payload to be signed:

1. The extension version byte, call, extension and extension implicit should be encoded;
2. The result of the encoding should then be hashed using the `BLAKE2_256` hasher;
3. The result of the hash should then be signed with the signature type specified in the extension definition.

```rust
// Step 1: encode the bytes
let encoded = (extension_version_byte, call, transaction_extension, transaction_extension_implicit).encode();
// Step 2: hash them
let payload = blake2_256(&encoded[..]);
// Step 3: sign the payload
let signature = keyring.sign(&payload[..]);
```
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This will most likely depend on which origin and kind of signature is used. For instance, ZK circuits might have different levels of proofs and might not be signing the whole payload. This should go to another RFC, which defines every particular VerifySignature-like extension.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, this is just an example on how to achieve Signed functionality in v5. I agree that the "chosen" way of providing this functionality in the actual runtimes should be a different RFC, dependent on this one.

However, there should be only one VerifySignature-like extension used to replicate this signed behavior. There is no need to have multiple extensions for this purpose. For custom origins, users can create infinite variations of an authorization extension and that doesn't need an RFC.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, but anyways this should go on a different RFC. I agree that VerifySignature-extension requires its own RFC, yes, but this has nothing to do with extrnsics v5 RFC.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Again, VerifySignature is just an example, and also very clearly labeled as an example in the latest revision of the RFC. Because of the explanation above and the fact that it's not obvious from the start what the extension pipeline should look like for v5 extrinsics, I think the example is useful.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the example is useful, and since it's also explicitly labeled as an example I see no reason to remove it


### Summary of changes in version 5

In order to minimize the number of changes to the extrinsic format version and also to help all
consumers downstream in the transition period between these extrinsic versions, we should:

- Remove the `Signed` variant starting with v5 extrinsics
- Add the `General` variant starting with v5 extrinsics
- Enable runtimes to support both v4 and v5 extrinsics

## Drawbacks

The metadata will have to accommodate two distinct extrinsic format versions at a given point in
time in order to provide the new functionality in a non-breaking way for users and tooling.
Comment on lines +120 to +121
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not a drawback IMO. Metadata v15 should show v4 and metadata v16 and ahead have a vector of extrinsic versions.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is something extra to support in the metadata for both the runtime and users, is this not a drawback?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it is both a drawback and an improvement - the new metadata support is an improvement, but having to do this enhancement to the metadata is a drawback to this RFC 😛

maybe add another line explicitly calling out that adding this metadata enhancement is ultimately a good thing that should be useful for potential future scenarios too

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added the explanation.


## Testing, Security, and Privacy

There is no impact on testing, security or privacy.

## Performance, Ergonomics, and Compatibility

This change makes the authorization through signatures configurable by runtime devs in version 5
extrinsics, as opposed to version 4 where the signing payload algorithm and signatures were
hardcoded. This moves the responsibility of ensuring proper authentication through
`TransactionExtension` to the runtime devs, but a sensible default which closely resembles the
present day behavior will be provided in `VerifySignature`.
Comment on lines +133 to +137
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not true. Signature schemes and addresses were configurable by runtime devs through rust generics. For example, Moonbeam uses only ECDSA signatures with EVM-like addresses.
Besides that, I wouldn't mention VerifySignature since it should be RFC-ed anyways.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Your statement about configurable signature schemes is true. However, I still consider my statement to be true because:

  • The signing payload generation algorithm was hardcoded; this is not the case anymore as any extension can take the inherited implication and add or subtract any data to it and mutate it in any way (such as hashing it - or not) before actually creating a signature.
  • There are now multiple ways of ending up with a Signed origin variant, with arbitrary logic in any TransactionExtension being able to authorize that origin; before, a user HAD to provide a transaction signed by a specific account.

All of this static logic is now moved to extensions. The extensions receive the inherited implication, the generation of which is still hardcoded and handled in this RFC, but is not in any way mandatory to be used in any signing scheme.

I'd agree though that the phrasing isn't clear, but I'm not sure how to improve it.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the phrasing is fine as it is. The point is to highlight the increase in configurability, which it does.


### Performance

There is no performance impact.

### Ergonomics

Tooling will have to adapt to be able to tell which authorization scheme is used by a particular
transaction by decoding the extension and checking which particular `TransactionExtension` in the pipeline is enabled to do the origin authorization. Previously, this was done by simply checking whether the transaction is signed or unsigned, as there was only one method of authentication.
georgepisaltu marked this conversation as resolved.
Show resolved Hide resolved

### Compatibility

As long as extrinsic version 4 is still exposed in the metadata when version 5 will be introduced,
the changes will not break existing infrastructure. This should give enough time for tooling to
support version 5 and to remove version 4 in the future.

## Prior Art and References

This is a result of the work in [Extrinsic
Horizon](https://github.com/paritytech/polkadot-sdk/issues/2415) and
[RFC99](https://github.com/polkadot-fellows/RFCs/blob/main/text/0099-transaction-extension-version.md).

## Unresolved Questions

There is no clear way to expose two different extrinsic versions in the current metadata framework.
A non-exhaustive list of options discussed so far:

1. Change the `ExtrinsicMetadata` trait to specify a list of versions instead of a single version.
2. Use the custom fields in the metadata to specify the details of the version 5.
3. Create a new trait similar to `ExtrinsicMetadata`, but for future versions of the extrinsic
format and add it to the metadata.
georgepisaltu marked this conversation as resolved.
Show resolved Hide resolved

## Future Directions and Related Material

Following this change, extrinsic version 5 will be introduced as part of the [Extrinsic
Horizon](https://github.com/paritytech/polkadot-sdk/issues/2415) effort, which will shape future
work.