Currently in EIP-1, except of Core EIPs there is no much mentioning criteria for allowing publication of a draft EIP or advancing an EIP’s status to final. This raises concern from some EIP contributors about quality, worthiness and whether there are sufficient interest or whether sufficient level of consensus has been reached.
In the interest of starting a discussion, I am creating this (probably dumb and full of flaws) proposal for criteria for advancing statuses of an EIP. Please see it as a straw-man and provide your feedback.
Standard Tracks
Category
Draft
Review
Last Call
Final
Core
Auto-merge
FEM discussion or 1st ref impl
Agreed on ACD & 2+ indenpendent impls
wait 14 days
Network
Auto-merge
FEM discussion or 1st ref impl
2+ independent impls
wait 14 days
Interface
Auto-merge
FEM discussion or 1st ref impl
3+ independent impls
wait 14 days
ERC
Auto-merge
FEM discussion and 1st ref impl
3+ independent impls
wait 14 days
Non-standard Tracks
Category
Draft
Review
Last Call
Final
Meta (binding)
FEM discussion
“rough consensus” (to be clarified)
wait 60 days
Informational(non-binding)
Auto-merge
Author discretion
FEM discussion
wait 60 days
Note: this was previously briefly discussed on EIP Editor Meeting #70 and another relevant thread
Given that a link to FEM discussion is a requirement for auto-merge, this metric needs to be higher. e.g. discussion on the draft EIP from 2 independent projects.
For ERC I feel that a minimum of 2 implementations is required to get to review. If a developing ERC only has one project supporting it then I don’t see it getting much decent input at the review stage.
Given that a link to FEM discussion is a requirement for auto-merge
In the above table, FEM discussion is not a requirement for auto-merge. What I meant to say is that for merging an EIP with Draft status, auto-merge means there is no particular Editor review needed.
For ERC I feel that a minimum of 2 implementations is required to get to review. If a developing ERC only has one project supporting it then I don’t see it getting much decent input at the review stage.
I can totally resonate with your sentiment. There is a spectrum of people’s preference. Some expressed to me that two ref-impls as a criteria for advancing to “Final” are too high a bar. Some (You @abcoathup for example) even think this requirement is too low a bar even for “Last Call” and hope to make it a bar as the bar for advancing to “Review”. Love to hear more people what they think.
I don’t think either of these are cut and dry drop in replacements for the governance here, but I think looking to what others have done and understanding why they’ve taken things that way will help to inform the direction we as a community wish to go.
One thing worth noting for both of these process first off is timelines. In both cases, when someone undertakes the process of producing a standard in either of these places I’d expect it to take roughly a year at least to go from first draft to completion. Usually it’s a multiyear effort though which can take 2 or 3 years. In fact, I’ve seen some take longer and it’s only because WG charters were set to expire that a first iteration of a standard was shipped.
So with that said, I think a good thing for us to set as a guiding consideration here is “How long do we expect it to take from an EIP first drafted to an EIP being shipped and multiple implementations are relying upon it?”
In my opinion, I think a good answer to this is between 6 to 18 months depending on the piece of work you want to ship and how many parties are working on this.
@kdenhartog
Thank you for providing the IETF, W3C process guide. I would look into them.
We discussed last time, and we generally feel that mandating a time is not the best way to address the merit or maturity of a pending EIP. Thus, I aim for more concrete criteria, e.g. independent reference implementations. But I am open to be convinced and could totally understand if there might be some very good reason for a timeline requirement.
I agree that we need something better than a timeline, so number of independent implementations seems to be the key measurement of a good spec to me as well. The main reason for a timeline is to prevent specs from becoming bloated or taking forever to ship because new people keep joining and requesting additional changes or adding additional requirements late in the process which can slow consensus down.
It’s worth pointing out though his point at 19:54 though which is that “Standards development doesn’t run out of time or money, but people in the WG run out of patience”. There’s a lot of wisdom in this statement because it suggests that if we don’t set arbitrary deadlines then we’re not likely to actually ship things and call something done. I’ve seen arbitrary deadlines work very well as a forcing function for creating consensus since most people would rather see something ship than nothing at all after they’ve spent a lot of time contributing to it’s development.
@gcolvin expressed in favor of having some implementations before ERC EIPs goes to Final
@poojaranjan expressed in favor of adding lower-bound time (30days) to advance to Final at least for Meta EIP
@SamWilsn@matt (lightclient) expressed objection to use implementations as a requirement for finalizing ERCs worrying it will add more hurdles for EIPs to get to final.
@SamWilsn also expressed that Editors might ask about worthiness of an EIP but “Non-of these editors will block” an EIP to advance to EIP even only author subjectively think this EIP is worthy.
@kdenhartog
If I understand you correctly, in your description of “timeline” you actually mean upper-bound of time allowed to advance to next status, right?
Yup, that’s a good way to describe it. If people operate in a faster time frame that’s a good indicator normally that things are non-controversial assuming there’s actually multiple implementations working on it.
Probably another thing that would be important here is that when EIPs hit particular milestones that there’s also a broad method of notifying people (basically people who don’t watch the EIP repo) that EIPs are hitting particular milestones and have ways to address objections during this process. E.g. Brave and Metamask are both on record for objecting to EIP-5749 (mine in the PR and FEM discussion and Metamasks in the FEM discussion ) but the EIP was still advanced forward anyways. That seems like a failure of the process if two wallets are actively against supporting an EIP directly related to their implementations, but the process doesn’t have any way to establish consensus.
Simply put, if wallets aren’t supporting this EIP it’s effectively DoA even if it does hit final. Especially if Metamask isn’t going to support it given their userbase. Furthermore, by publishing this EIP it legitimizes it in a way that creates more web compatibility issues for DApps like Opensea which may now have to support an additional property if they want their users to be able to use MEW wallet. My goal is to avoid these types of images showing up on DApps.
Generally in favor of creating a guideline (but not a criteria) for EIP to advance status
Generally in favor of seeing implementations for last call and final as a signal for community interest
Mentioned that having multiple implementations might not sufficiently demonstrate independence between them but lacking implementations could be seen as a signal of lack of community interest
@xinbenlv@matt@SamWilsn participated in discussion. Correct me if my understanding is wrong.
