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

Address metadata registry #926

Merged
merged 4 commits into from
Mar 29, 2018
Merged
Changes from 1 commit
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
72 changes: 72 additions & 0 deletions EIPS/EIP-X-metadata-nickjohnson.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,72 @@
## Preamble

EIP: <to be assigned>
Title: Address metadata registry
Author: Nick Johnson <nick@ethereum.org>
Type: Standard track
Category: ERC
Status: Draft
Created: 2018-03-12

## Abstract
This EIP specifies a registry for address metadata, permitting both contracts and external accounts to supply metadata about themselves to onchain and offchain callers. This permits use-cases such as generalised authorisations, providing token acceptance settings, and claims registries.

## Motivation
An increasing set of use cases require storage of metadata associated with an address; see for instance EIP 777 and EIP 780, and the ENS reverse registry in EIP 181. Presently each use-case defines its own specialised registry. To prevent a proliferation of special-purpose registry contracts, we instead propose a single standardised registry using an extendable architecture that allows future standards to implement their own metadata standards.

## Specification
The metadata registry has the following interface:
```
interface AddressMetadataRegistry {
function provider(address target) view returns(address);
function setProvider(address _provider);
}
```

`setProvider` specifies the metadata registry to be associated with the caller's address, while `provider` returns the address of the metadata registry for the supplied address.

The metadata registry will be compiled with an agreed-upon version of Solidity and deployed using the trustless deployment mechanism to a fixed address that can be replicated across all chains.

## Provider specification

Providers may implement any subset of the metadata record types specified here. Where a record types specification requires a provider to provide multiple functions, the provider MUST implement either all or none of them. Providers MUST throw if called with an unsupported function ID.
Copy link
Contributor

Choose a reason for hiding this comment

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

For clarity, recommend including the Solidity code for this:

function () { revert; }

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Solidity implements this automatically now.


Providers have one mandatory function:

```
function supportsInterface(bytes4 interfaceID) constant returns (bool)
```

The `supportsInterface` function is documented in [EIP 165](https://github.com/ethereum/EIPs/issues/165), and returns true if the provider implements the interface specified by the provided 4 byte identifier. An interface identifier consists of the XOR of the function signature hashes of the functions provided by that interface; in the degenerate case of single-function interfaces, it is simply equal to the signature hash of that function. If a provider returns `true` for `supportsInterface()`, it must implement the functions specified in that interface.
Copy link
Contributor

Choose a reason for hiding this comment

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

What is the motivation behind using XOR of the signature IDs rather than a hash of all of the signatures? While I suspect someone can implement the XOR without too much difficulty, keccak hashing of things in the Solidity ecosystem is a well understood problem as it is used everywhere while XOR of first 4 bytes of a keccak hash of a thing is significantly less understood by the average developer.

The following is what the hash might look like, which is pretty familiar to any Ethereum developer at this point and can be done in Solidity or in JavaScript (most common dapp development language).

keccak256("function foo(uint256,bool)", "function bar()", "function zip(string,bytes)")

Copy link
Contributor Author

Choose a reason for hiding this comment

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

See EIP 165 for justification. I don't think this is a good place to rehash that.

Copy link
Contributor

Choose a reason for hiding this comment

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

Glanced through that, there was a reference to a Gitter conversation where bytes4 XOR was chosen over bytes32 hash but no details of the contents of that conversation were provided. From my reading of #165, I am not at all convinced that XOR of function IDs is superior to hash of function signatures.

Was there a particular comment that you feel captures the argument over there? It is possible I missed a subtle argument as I was skimming the comments rather than analyzing them in depth.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I really don't want to get derailed on arguing the merits of an unrelated standard in this EIP. Can you open a discussion on 165 instead? It's unlikely we can make a change at this point, though, since it's already widely deployed (eg, on ENS for one).

Copy link
Contributor

Choose a reason for hiding this comment

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

Yeah, now that I have read EIP 165 and realize that this section is just a watered down summary of 165 I no longer want to press the bytes32 vs XOR thing here.

Copy link
Contributor

Choose a reason for hiding this comment

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

This should link to the draft of EIP 165, not the issue: https://github.com/ethereum/EIPs/blob/master/EIPS/eip-165.md

Copy link
Contributor

Choose a reason for hiding this comment

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

After reading EIP 165 (draft rather than the GitHub issue, which was much harder to follow), I am now confused why this entire section exists. It appears to just be re-hashing what EIP 165 says, but with less context, less detail and fewer examples. Why not just say that providers are defined by EIP 165 and leave it at that?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I thought a brief description would be useful so that someone who only wants to implement it doesn't need to read all of 165, too.

Copy link
Contributor

Choose a reason for hiding this comment

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

I need to step away and think a bit more to avoid curve fitting my personal experience, but my gut right now is telling me that the fact that there was a bunch of text rather than just a link resulted in me ignoring the link and reading just the text, which unfortunately is lacking in the detail I would expect from an EIP.


`supportsInterface` must always return true for `0x01ffc9a7`, which is the interface ID of `supportsInterface` itself.
Copy link
Contributor

Choose a reason for hiding this comment

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

What is the point of this? At first blush it feels like it is just a clever little thing to chuckle at in the spec, but not something that anyone would ever need for anything. I'm guessing the thought is that you can call this to check if supportsInterface is available at all (with the expectation that it will return false by default) but I believe there is currently a "bug" in Solidity that makes the default return garbage (which evaluates to false 1/2^256 times). Even if this is "fixed" (unclear if that is planned or not), it feels weird probing a contract like this.

Is there some other reason for this MUST?

Copy link
Contributor

Choose a reason for hiding this comment

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

Also, recommend changing must to **MUST**.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Also see 165 and 137; the idea is that it lets a generic caller determine if the contract implements supportsInterface in the first place.

Copy link
Contributor

Choose a reason for hiding this comment

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

As mentioned over in #165 (and I didn't see a rebuttal, though I may have missed it) older Solidity (based on discussion there it sounds like this has since been fixed) contracts will return garbage data by default (meaning true or false). Thus if someone calls supportsInterface on such a contract they may (likely) get back true even though it doesn't actually support the interface. One recommendation was to have both a positive and a negative check function to address this.

Copy link
Contributor

Choose a reason for hiding this comment

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

It appears that 165 addresses this problem with:

true when interfaceID is 0x01ffc9a7 (EIP165 interface)
false when interfaceID is 0xffffffff

Copy link
Contributor Author

Choose a reason for hiding this comment

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

As mentioned over in #165 (and I didn't see a rebuttal, though I may have missed it) older Solidity (based on discussion there it sounds like this has since been fixed) contracts will return garbage data by default (meaning true or false).

Only if they implement a fallback that doesn't throw - which this standard explicitly prohibits.


The first argument to all provider functions MUST be the address being queried; this facilitates the creation of multi-user provider contracts.

Currently standardised provider interfaces are specified in the table below.

| Interface name | Interface hash | Specification |
| --- | --- | --- |

EIPs may define new interfaces to be added to this registry.

## Rationale
There are two obvious approaches for a generic metadata registry: the indirection approach employed here, or a generalised key/value store. While indirection incurs the cost of an additional contract call, and requires providers to change over time, it also provides for significantly enhanced flexibility over a key/value store; for that reason we selected this approach.

## Backwards Compatibility
There are no backwards compatibility concerns.

## Implementation
The canonical implementation of the metadata registry is as follows:
```
contract AddressMetadataRegistry {
mapping(address=>address) public provider;

function setProvider(address _provider) {
provider[msg.sender] = _provider;
}
}
```
Copy link
Contributor

Choose a reason for hiding this comment

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

Recommend including an example provider as well that at least implements the required supportsInfercae code (and shows how to do the function signature hashing in Solidity).

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Agreed, though callers don't need to do the function hashing, just embed the appropriate constants (unless we can trust solidity to optimise those out, even if optimisation is turned off?)

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 there is value in showing a user how to do the XOR, hashing and byte chopping in code, and Solidity is the language that we can be certain all readers of these understand.

Copy link
Contributor Author

@Arachnid Arachnid Mar 12, 2018

Choose a reason for hiding this comment

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

If Solidity doesn't optimise-away the hashing at compile-time, though, this is going to be a significant waste of gas.

Edit: And in any case this is likely something that should be demoed in 165.

Copy link
Contributor

Choose a reason for hiding this comment

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

I would be OK (though not quite as satisfied) with an example written in JS if you think that is more reasonable. Solidity is just the "common language". I would not recommend people actually do the hashing in Solidity with each contract 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.

Again, I don't think this is the place to do it - this is just a user of EIP165; any examples should probably be there.

Copy link
Contributor

Choose a reason for hiding this comment

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

As a developer who didn't know of EIP 165's prior to reading this EIP, I was surprised to find it lacking an example for the "other side". If you think this EIP should depend on 165, then I recommend including such a dependency in the EIP and indicating how the two are connected to each other. In general, if an EIP doesn't reference any other EIPs, then I assume it can be consumed on its own (which is what I tried to do when reading this EIP) and it seems that your intent is that this EIP is not consumed on its own.

Copy link
Contributor

Choose a reason for hiding this comment

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

And of course after saying that, I re-read this EIP and found it did mention EIP 165...


## Copyright
Copyright and related rights waived via [CC0](https://creativecommons.org/publicdomain/zero/1.0/).