feat(pattern): privacy L2s comparison #71

Meyanis95 · 2026-01-28T10:25:20Z

Summary

Added L2 Privacy Evaluation Framework pattern for comparing privacy-preserving L2 solutions
Updated RFP for Living Benchmark Dashboard
Added .gitignore (node_modules, package-lock.json, editor files, Vale cache)
Synced Vale write-good style (required by .vale.ini)

Test plan

npm run validate passes (0 errors)
npm run check-terms runs (warnings in other files)
npm run lint:vale passes (0 errors)

@oskarth
Vale setup: Running vale sync downloaded the write-good package to .vale/styles/. Is this the intended setup, or should these styles be excluded from the repo (added to .gitignore) and installed separately by contributors? Currently they're committed (~1.2K lines).

oskarth · 2026-01-28T11:32:18Z

CI Fix Summary

The CI was failing on the Prose Quality (Vale) check due to two issues:

1. CHANGELOG.md Spelling Errors

The .vale.ini config applied only the base Vale style to CHANGELOG.md, which doesn't include the IPTF vocabulary. Domain terms like "Nullifiers", "Ethereum", "ZKsync", etc. triggered spelling errors.

Fix: Changed BasedOnStyles = Vale to BasedOnStyles = IPTF for CHANGELOG.md section.

2. Reviewdog Annotation Limit

The Vale action uses reviewdog to post annotations, but GitHub has a limit on annotations per workflow run. When Vale produces many warnings (even non-blocking ones), reviewdog fails with "Too many results (annotations) in diff" - even with fail_on_error: false.

Fix: Changed reporter from default github-pr-annotations to github-check, which has a higher limit and handles large result sets better. See errata-ai/vale-action#89 for context.

Why It Passed Locally But Failed in CI

Locally, npm run lint:vale runs Vale directly without reviewdog, so it only shows warnings/errors without hitting annotation limits. In CI, the vale-action wraps Vale with reviewdog for PR annotations, which has stricter limits.

Commits:

e811197 - fix(ci): use IPTF style for CHANGELOG Vale checks
9b9c8e5 - fix(ci): use github-check reporter for Vale action

Vale's base spelling checker doesn't use the IPTF vocabulary. Switch CHANGELOG.md to IPTF style to recognize domain terms.

The default github-pr-annotations reporter hits GitHub's annotation limit when Vale produces many warnings, causing reviewdog to fail even with fail_on_error: false. The github-check reporter has a higher limit and handles large result sets better.

Meyanis95 · 2026-01-29T08:17:51Z

Ping @oskarth for review

oskarth · 2026-01-29T08:29:17Z

.vale/styles/write-good/Cliches.yml

+  - think outside of the box
+  - third time's the charm
+  - this day and age
+  - this hurts me worse than it hurts you


cc @mojalil

oskarth · 2026-01-29T08:29:34Z

.vale/styles/write-good/Cliches.yml

+  - no skin off my back
+  - no stone unturned
+  - no time like the present
+  - no use crying over spilled milk


this list man lmao

oskarth · 2026-01-29T08:31:48Z

patterns/pattern-l2-privacy-evaluation.md

+## Protocol
+
+1. **Define workload**: Use [Simple Value Transfer](#simple-value-transfer) as baseline
+2. **Collect metrics**: Request L2 teams fill in the three evaluation tables


Why is collecting metrics part of the protocol here? Just making sure this actually matches intention

This describes the process of coming up with a privacy/throughput comparative table.
Not sure if it's clear.

Is the "process" what we want to capture? As opposed to the actual comparative table itself? I thought the latter would be more useful, and then just have it well-sourced and somewhat up to date?

Or do you expect many people to want the process itself? Almost seems more like a CLAUDE skill or so?

What we want to capture are both the framework and the results. Both wouldn't fit as a pattern to be semantically correct, but that was the only place where it'd fit. The PRD mentioned domains but I assume it was wrong.
The goal is still to have the analysis with L2 teams provided metrics, but as of today nobody answered, so the tables here are LLM placeholders.

So the question is should create a new category in the repo to capture this? Or make it fit the pattern even though it's not a direct match?

oskarth · 2026-01-29T08:32:49Z

patterns/pattern-l2-privacy-evaluation.md

+| **Scroll Cloak** |     Yes      |    Access control    | Operator + regulator access |
+| **EY Nightfall** |     Yes      |         Yes          | Enterprise audit trail      |
+
+_Sources: Protocol documentation, L2Beat, academic papers. Last updated: 2026-01-27_


Have all these been reviewed by respective projects? Might be good to add sources or so, including if it is asking Core Devs or so in Jan, 2026. Ideally post to docs or something as source.

No, this is only publicly available information, mainly from each L2 documentation.
Yes, I can add sources.

Added sources and disclaimer to the doc.

oskarth · 2026-01-30T02:07:51Z

Vale setup: Running vale sync downloaded the write-good package to .vale/styles/. Is this the intended setup, or should these styles be excluded from the repo (added to .gitignore) and installed separately by contributors? Currently they're committed (~1.2K lines).

I don't know. I didn't see this before. It does seem a bit noisy? Maybe we could add to gitignore, and if it seems valuable to have in code base we can add in separate PR later?

Meyanis95 · 2026-01-30T08:27:32Z

Vale setup: Running vale sync downloaded the write-good package to .vale/styles/. Is this the intended setup, or should these styles be excluded from the repo (added to .gitignore) and installed separately by contributors? Currently they're committed (~1.2K lines).

I don't know. I didn't see this before. It does seem a bit noisy? Maybe we could add to gitignore, and if it seems valuable to have in code base we can add in separate PR later?

Tbf I just ran the specified command and Vale automatically added as part of the setup.
Happy to prune if it wasn't intentional to have the good-write checks.

Meyanis95 requested a review from oskarth January 28, 2026 10:28

Meyanis95 changed the title ~~feat: pr 011 -- l2 privacy comparison~~ feat(pattern): l2 privacy comparison Jan 28, 2026

Meyanis95 changed the title ~~feat(pattern): l2 privacy comparison~~ feat(pattern): privacy l2s comparison Jan 28, 2026

Meyanis95 changed the title ~~feat(pattern): privacy l2s comparison~~ feat(pattern): privacy L2s comparison Jan 28, 2026

Meyanis95 and others added 5 commits January 28, 2026 12:42

feat: Privacy L2 analysis framework + MvP

6c62e55

feat(chore): added gitignore

b706dc8

feat(chore): lint + vale checks

6db577f

fix(ci): use IPTF style for CHANGELOG Vale checks

4f55ce7

Vale's base spelling checker doesn't use the IPTF vocabulary. Switch CHANGELOG.md to IPTF style to recognize domain terms.

Meyanis95 force-pushed the pr-011-l2-privacy-comparison branch from 9b9c8e5 to e496e31 Compare January 28, 2026 11:43

This was linked to issues Jan 29, 2026

feat(domains): L2 privacy comparison document (PR-011) #62

Open

feat: map performance and trust assumptions of privacy L2 designs #27

Open

oskarth reviewed Jan 29, 2026

View reviewed changes

Meyanis95 and others added 2 commits January 29, 2026 09:58

feat: added sources

2099514

Merge branch 'master' into pr-011-l2-privacy-comparison

a64d5d1

Meyanis95 requested a review from oskarth January 29, 2026 11:57

feat(pattern): privacy L2s comparison #71

Are you sure you want to change the base?

feat(pattern): privacy L2s comparison #71

Uh oh!

Conversation

Meyanis95 commented Jan 28, 2026

Summary

Test plan

Uh oh!

oskarth commented Jan 28, 2026

CI Fix Summary

1. CHANGELOG.md Spelling Errors

2. Reviewdog Annotation Limit

Why It Passed Locally But Failed in CI

Uh oh!

Meyanis95 commented Jan 29, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

oskarth commented Jan 30, 2026

Uh oh!

Meyanis95 commented Jan 30, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants