You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Paralegal research for attorney review; not legal advice. This issue creates a license-agnostic research corpus. It does not recommend AGPL, GPL, MPL, the FastLED draft, or any other license, and it does not authorize enforcement activity or changes to LICENSE.
Context
The repository now has a useful but narrow starting point: PR-0001 addresses a U.S./California AGPL-style remedies hypothetical, supported by eight authority cards. It is not yet broad enough to power a general paralegal AI or give reviewing counsel a reliable view of related reciprocal and source-availability licenses.
The missing substrate includes a license-family comparison, a systematic case/docket census, non-U.S. enforcement, doctrine-specific synthesis, defenses, standing, source/derivative-work scope, registration, damages, evidentiary proof, procedure, and a precise record of what relief was requested versus actually ordered.
Recent source checks confirm additional material that is not in the corpus, including the XimpleWare/Versata litigation, Neo4j/AGPL litigation, French Entr'Ouvert/Orange decisions, EU software-license doctrine, and German GPL enforcement decisions. These must be researched from primary sources and recorded with procedural posture and later history.
Proposal
Build an indexed, license-agnostic legal research corpus under paralegal-research/ for use by a paralegal AI and final review by human open-source counsel.
1. License-family corpus
Create reusable license cards and a comparison matrix for, at minimum:
GPL-2.0 and GPL-3.0
LGPL-2.1 and LGPL-3.0
AGPL-3.0
MPL-2.0
EPL-2.0
CDDL-1.0
EUPL-1.2
OSL-3.0
RPL-1.5
SSPL-1.0
BUSL-1.1
Elastic License 2.0
Commons Clause
representative PolyForm and Fair Source/FSL terms
the FastLED Reciprocal License draft only as one comparison target
For each, record approval/status accurately and distinguish open-source licenses from source-available licenses. Compare trigger, covered-work boundary, disclosure recipient, disclosure timing, duration, cure/reinstatement, commercial alternative, patent terms, compatibility, governing law/forum, standing language, and known judicial treatment.
2. Case and enforcement census
Systematically locate and classify judicial dispositions, dockets, complaints, orders, judgments, and publicly documented settlements involving reciprocal/open-source/source-availability license enforcement.
The initial U.S. set must include and expand beyond Jacobsen, MDY, Artifex/Hancom, BusyBox enforcement, SFC/Vizio, XimpleWare/Versata/Ameriprise, Neo4j/PureThink/Graph Foundation, Wallace, and Planetary Motion. The comparative set must cover available primary authority from the EU, France, Germany, and other jurisdictions surfaced by the search, including Entr'Ouvert/Orange, relevant CJEU software-license decisions, and the German GPL enforcement line.
Each authority card must separately identify:
direct primary source and stable citation;
court, jurisdiction, date, docket, later history, and precedential weight;
license/version and conduct at issue;
claimant and standing theory;
allegations and requested relief;
procedural ruling and actual holding;
relief actually awarded;
settlement terms, if public, explicitly labeled non-precedential;
what the authority does not establish;
relevance to other license families.
3. Doctrine synthesis
Create separate canonical matter memos covering:
public-license formation and assent;
copyright condition versus contractual covenant;
federal preemption and copyright nexus;
copyright ownership, assignments, contributor standing, and third-party beneficiaries;
covered/derivative-work, linking, aggregation, corresponding-source, and distribution boundaries;
network-use and deployment triggers;
termination, notice, cure, reinstatement, and past liability;
injunctions, specific performance/source disclosure, impoundment, declaratory relief, and contempt;
actual damages, profits, statutory damages, fees, commercial-license comparables, and apportionment;
registration, limitations periods, accrual, and continuing conduct;
defenses including authorization, waiver, estoppel, misuse, unclean hands, impossibility, mootness, and voluntary cessation;
evidence preservation, binary/source comparison, version attribution, public-source availability, and chain of custody;
pleading, jurisdiction, venue, choice of law, discovery, and protective orders;
supply-chain actors, downstream recipients, resellers, manufacturers, and secondary liability;
settlement/compliance structures versus judicially available remedies;
comparative and cross-border enforcement.
4. AI retrieval and quality controls
Extend the index and agent documentation so an AI can answer by proposition rather than keyword alone. It must route by actor, conduct, jurisdiction, doctrine/remedy, license/version, authority posture, and verification date.
Research agents may create only unique assigned matter drafts. A coordinator must deduplicate authorities, create shared authority cards, update INDEX.md, and resolve cross-links so concurrent agents do not overwrite shared files or create multiple cards for the same decision.
Acceptance criteria
Existing PR-0001 and AUTH-0001 through AUTH-0008 are preserved, corrected where primary sources require it, and reused rather than duplicated.
A documented search log identifies databases, search terms, jurisdictions, date searched, inclusion/exclusion rules, unresolved leads, and negative-result searches.
The minimum license set has official-text cards and a normalized comparison matrix.
Every located case used for a proposition has an authority card with posture, weight, requested relief, actual relief, limitations, later history, direct source, and verification date.
Complaints, interlocutory rulings, default judgments, final contested judgments, settlements, and commentary are never conflated.
The doctrine topics above have scoped matter memos, explicit assumptions, contrary authority, jurisdiction limits, and open questions for counsel.
U.S. research distinguishes federal copyright rules from state-specific contract and equitable-remedy rules.
Comparative research uses official court or government sources where available and records translation provenance.
The corpus contains no statement that source publication or purchase of a commercial license is an automatic remedy unless supported by final applicable authority.
Open-source and source-available classifications are accurately separated.
All matter IDs and authority IDs are unique, indexed, cross-linked, and machine-searchable.
Local Markdown links resolve, git diff --check passes, and all records remain marked for attorney review unless counsel expressly approves them.
No change is made to LICENSE as part of this research issue.
A final coverage report lists researched propositions, confidence, unresolved splits/gaps, stale-source triggers, and the questions requiring counsel's judgment.
Decisions
Scope is license-agnostic. FastLED is a comparison target, not the organizing principle, because the research must transfer to related licenses and future agent skills.
Breadth is systematic, not literally infinite. Completion is measured by the defined license set, doctrine set, jurisdiction plan, search log, and documented residual gaps; this makes the work auditable and repeatable.
Primary authority controls. Secondary sources may identify leads but cannot establish a holding or remedy when a court document is available.
Jurisdictions begin with U.S. federal/state law plus EU, France, and Germany. Other countries are added when material primary authority is found; a superficial worldwide survey is excluded.
The corpus precedes skill packaging. This issue builds the trusted source layer and retrieval protocol; packaging it as a reusable Codex/paralegal skill is a later, separately reviewed task.
Terra agents research bounded topics; one coordinator integrates. This preserves parallel speed without shared-index collisions or duplicate authority cards.
Attorney review remains a separate status. Research completion never implies legal approval.
Open questions
Which defendant forums and non-U.S. markets should counsel prioritize after the general corpus is complete?
Which licensed releases and ownership records should be used for a later project-specific standing and registration audit?
Should counsel require certified translations for any non-English decision that materially affects final advice?
Important
Paralegal research for attorney review; not legal advice. This issue creates a license-agnostic research corpus. It does not recommend AGPL, GPL, MPL, the FastLED draft, or any other license, and it does not authorize enforcement activity or changes to
LICENSE.Context
The repository now has a useful but narrow starting point: PR-0001 addresses a U.S./California AGPL-style remedies hypothetical, supported by eight authority cards. It is not yet broad enough to power a general paralegal AI or give reviewing counsel a reliable view of related reciprocal and source-availability licenses.
The missing substrate includes a license-family comparison, a systematic case/docket census, non-U.S. enforcement, doctrine-specific synthesis, defenses, standing, source/derivative-work scope, registration, damages, evidentiary proof, procedure, and a precise record of what relief was requested versus actually ordered.
Recent source checks confirm additional material that is not in the corpus, including the XimpleWare/Versata litigation, Neo4j/AGPL litigation, French Entr'Ouvert/Orange decisions, EU software-license doctrine, and German GPL enforcement decisions. These must be researched from primary sources and recorded with procedural posture and later history.
Proposal
Build an indexed, license-agnostic legal research corpus under
paralegal-research/for use by a paralegal AI and final review by human open-source counsel.1. License-family corpus
Create reusable license cards and a comparison matrix for, at minimum:
For each, record approval/status accurately and distinguish open-source licenses from source-available licenses. Compare trigger, covered-work boundary, disclosure recipient, disclosure timing, duration, cure/reinstatement, commercial alternative, patent terms, compatibility, governing law/forum, standing language, and known judicial treatment.
2. Case and enforcement census
Systematically locate and classify judicial dispositions, dockets, complaints, orders, judgments, and publicly documented settlements involving reciprocal/open-source/source-availability license enforcement.
The initial U.S. set must include and expand beyond Jacobsen, MDY, Artifex/Hancom, BusyBox enforcement, SFC/Vizio, XimpleWare/Versata/Ameriprise, Neo4j/PureThink/Graph Foundation, Wallace, and Planetary Motion. The comparative set must cover available primary authority from the EU, France, Germany, and other jurisdictions surfaced by the search, including Entr'Ouvert/Orange, relevant CJEU software-license decisions, and the German GPL enforcement line.
Each authority card must separately identify:
3. Doctrine synthesis
Create separate canonical matter memos covering:
4. AI retrieval and quality controls
Extend the index and agent documentation so an AI can answer by proposition rather than keyword alone. It must route by actor, conduct, jurisdiction, doctrine/remedy, license/version, authority posture, and verification date.
Research agents may create only unique assigned matter drafts. A coordinator must deduplicate authorities, create shared authority cards, update
INDEX.md, and resolve cross-links so concurrent agents do not overwrite shared files or create multiple cards for the same decision.Acceptance criteria
git diff --checkpasses, and all records remain marked for attorney review unless counsel expressly approves them.LICENSEas part of this research issue.Decisions
Open questions
Related issues