[EP3.7] [Social] Approval of ENS Name Normalization Standard (ENSIP-15)
This is a vote to approve ENSIP-15: Normalization Standard.
- Normalization isn’t enforced on-chain.
- There is no code for the DAO to execute.
- Approval for ENSIP-15 should be confirmed through a social vote.
- Since ENSIP-1 (originally EIP-137) was finalized in 2016, Unicode has evolved from version 8.0.0 to 15.0.0 and incorporated many new characters, including complex emoji sequences.
- ENSIP-1 does not state the version of Unicode.
- ENSIP-1 implies but does not state an explicit flavor of IDNA processing.
UTS-46 is insufficient to normalize emoji sequences. Correct emoji processing is only possible with UTS-51.
- Validation tests are needed to ensure implementation compliance.
- The success of ENS has encouraged spoofing via the following techniques:
- Insertion of zero-width characters.
- Using names which normalize differently between algorithms.
- Using names which appear differently between applications and devices.
- Substitution of confusable (look-alike) characters.
- Mixing incompatible scripts.
- Replace ENSIP-1 § Name Syntax “UTS-46 algorithm” with link to ENSIP-15.
- Agree to normalize names according to ENSIP-15 for a safer end-user experience.
- Names visible to the end-user should be beautified for a more consistent appearance.
- Example: These labels are the same:
This vote is a single choice vote. You may vote for one of the following options:
By voting For this proposal, you are voting in favor of approving ENSIP-15.
Just wanted to say @raffy you’ve done amazing work on Normalization
This looks great to me. Technically the new app is already using this too!
Does a new ENSIP need a Snapshot vote? I see other newer ENSIPs too, but I don’t think we’ve ever voted on them explicitly, as they don’t affect the core ENS protocol.
In general no. Given this is a significant change in how ENS names resolve, which affects existing names, I don’t think there’s any harm in getting formal DAO approval for it.
Just checking, we’re all good to go with this right? Is anyone opposed to putting this up to vote on Snapshot?
I ask because I think OpenSea is waiting for “official” approval before they will do a full reindex (so that previously delisted names can be relisted)
Also it looks like Metamask is using eth-ens-namehash, does that need to be updated too? And then Metamask would need to update to that newer version?
There is an eth-ens-namehash branch that uses the latest spec.
However, there were a few issues with ens-normalize.js deployment related to npm and Typescript that should probably be incorporated into this fork for maximum compatibility.
Additionally, eth-ens-namehash was commonjs and the new build is esm. To reduce any upgrade friction, it should probably have a commonjs build.
Let me make those changes and update this thread.
I think () this eth-ens-namehash branch is now a zero-change drop-in replacement for the old library.
npm pack and import into a new node project, it functions correctly as CJS or ESM.
Now that ENSIP 15 is submitted as a draft, this can proceed! @raffy, can you create a new EP instructions here and submit it as a draft? We’ll get it up for a vote on Snapshot.
Hey Raffy, can you please edit the original post above with the latest version of the draft proposal.
I would suggest posting the proposal to Snapshot on Monday, June 12.
It’s so exciting to see this Draft PR !
Also nice to see this added to the specification:
Names visible to the end-user should be beautified for a more consistent appearance.
What do you think about adding something like the following related idea to the specification:
- If the primary name of an account is not normalized then it should be treated as if the account has no primary name set.
There’s probably a better way to communicate that idea. It’s quite late here right now as I write this. The bigger goal behind this suggestion is to correct for cases where we see people set their primary name as a non-normalized name.
For example: Let’s say a dapp shows “ABC.eth” as a non-normalized primary name for account 0x123 to users in their UIs. Then someone copy-pastes that non-normalized name into a properly implemented text field that performs ens-normalize. The result is that the dapp actually interacts with “abc.eth” which might resolve to a different account 0x789 than the account the user thought they would be interacting with.
In practice, this would mean that the various libraries that implement the primary name reverse lookup logic would need to also perform an ens-normalize check, in addition to the other checks performed such as verifying forward resolution, etc…
Yeah, I agree with this, I was just unsure if it should be part of normalization or part of an additional document that covers best practices.
I included beautification as part of the spec because it requires normalization logic to apply it, but it’s really an UX thing that happens after normalization. For example, I have an demo page which demonstrates some of the issues (unpredictable emoji presentation between devices: compare mobile vs desktop, and some normalized emoji acquire text-coloring whereas beautified emoji do not: eg. make a text-style heart emoji appear blue vs blue heart emoji.)
I explored the invalid-primary case in my resolver demo: bld.eth. I also provide hints regarding address/owner/controller/primary sameness/mismatches.
There is another issue when a primary is set to an address-like hex string. This matters when a wallet is shown as one entity (ENS name or address) rather than both values. Across my various ENS experiments, I try to differentiate these two situations:
And there’s a minor issue in clearly differentiating subdomains:
Lastly, there are a slew of Unicode-related things, like normalized substring/fragment search or how to fix/cure a name, strip emoji, strip null labels, etc. that I intend to add to the technical docs.
I’m not suggesting to use any of my ideas, I’m not a designer and like simple hex colors
We could ask: What would be the best way to support the case that libraries implementing primary name reverse lookups update their logic to be more safe with this proposed ens-normalize check?
Whatever helps to protect the community from potentially nasty unexpected edge cases seems best.
My suggestion would be to take this opportunity to include this idea in ENSIP 15 as I believe it’s an important detail in achieving consistent UX for the community in relation to normalized names.
Feedback always welcome of course
The case of someone setting their primary name to an address-like hex string is quite interesting My first impression is that while it is a bit odd, it should all pass the test of: If a user copy/pastes this name into a properly implemented dapp that uses ens-normalize, will they ultimately resolve that properly? And it seems to me that they should.
For the issue of clearly differentiating subdomains, I’m not sure I understand this one? abc.def.eth seems clearly differentiated from abcdef.eth for me.
All the ideas you suggested adding to technical docs sounds great
Edited above to be more explicit in communicating a suggestion.
Okay, I think I did it — Snapshot
I’d like to push a few minor formatting updates to my ENSIP-15 docs. I’d also like to include the minor addendum to ENSIP-1 which points to ENSIP-15. Can I just PR these changes to the docs repo?
I did not include Lightwalker’s suggestions as I think they should go in a different location. What’s the proper protocol for adding these suggestions? Figure out what edits to make with a thread and then PR the changes, or is something more formal required?