Compare, honestly
a. Everything you keep
Common to both models. Nothing in this column is traded away by adopting.
Kept — identical under both models
| Dimension | Open source | Purpose Source |
|---|---|---|
| Public source code, readable by anyone | yes | yes |
| Anyone can inspect and learn from the code | yes | yes |
| Fork and clone; modify and adapt | yes | yes |
| Community issues, discussions, pull requests — no copyright assignment | yes | yes |
| Transparent development history; community-driven change | yes | yes |
| Free for individuals, students, hobbyists — no account, no registration | yes | yes |
| Free for nonprofits and organisations below the published threshold | yes | yes |
| No charge to contribute | yes | yes |
| Security auditability — inspect, scan, verify builds | yes | yes |
| No vendor lock-in — self-patch, self-build, self-host | yes | yes |
| Contributor reputation — public git history, portfolio value | yes | yes |
| Maintainer controls roadmap and merges | yes | yes |
| Zero-effort adoption and exit — one committed licence file each way | yes | yes |
| Old releases keep their licence forever — nothing is ever pulled back | yes | yes |
| Abandonment insurance — the code survives the maintainer | yes | yes |
| Education, research, and reference value | yes | yes |
| Eventual permissive commons | immediate | each version becomes Apache-2.0 four years after its release, in the text |
| Free commercial use by any organisation regardless of size | yes | intentionally different — that is the point |
b. What only Purpose Source adds
Each row is a mechanism that exists or is specified, not a projection. Nothing here is claimed as a result yet — no money has moved.
Added — with the reason it cannot come from a permissive licence alone
| Dimension | How it works, and why the other model cannot |
|---|---|
| Large-organisation use automatically creates public benefit | Fees route to vetted funds after published, capped, audited costs. |
| Charity funding built into licensing; the repository generates benefit passively | A permissively-licensed repository produces nothing for anyone by existing. |
| Existing work generates new future benefit | A mature codebase converts forward without a rewrite; past releases are untouched. |
| Contributor Impact Shares and a vote on causes | Directing, never receiving. Materiality-gated display, no leaderboards. |
| Verifiable contributor and steward certificates | Signed, machine-verifiable, with the algorithm version published. |
| A public, append-only allocation ledger | Nothing to hide because everything recorded is published. |
| A corporate compliance credential — one Pass, one review, everything covered | Machine-verifiable, with amnesty covenants on purchase. A permissive licence has no counterparty who could offer one. |
| Procurement-grade term certainty | Per-version vesting, a steward-lapse backstop, and a kill protocol that leaves payers whole. |
| Fee invoiced as a software licensing fee | Typically better-treated than a cross-border donation — a qualified comparison, never a guarantee; obtain your own advice. |
| Gratis, public fee waivers per named organisation | Selective generosity a permissive licence cannot express, and cannot be sold — selling one is a delisting offence. |
| Central charity vetting | The maintainer never operates a charity programme; curated category funds via an established intermediary. |
| Declared-usage demand signal from paying organisations | Adoption evidence that permissive maintainers never get. |
| An anti-capture constitution | Capped costs, no distributable private profit, gratis waivers — as never-reopen clauses. |
| A cross-project identity | Design intent: one profile aggregates a contributor's impact across every registered repository they touch, a maintainer's across their projects, a company account's across its entitlements. A permissive licence has no registry to aggregate across. |
c. Where open source keeps an edge — and the honest answer
The costs of adopting this licence, listed by us. Each answer is what we can actually say, not what would sound best.
Traded — the real costs, and the honest answer to each
| Where open source keeps an edge | The honest answer |
|---|---|
| The OSI label and the words themselves | Never claimed. This is not open source, and every version literally becomes open source four years after its release. |
| Universal scanner allowlists; frictionless corporate ingestion | An SPDX identifier is requested at launch, with one canonical zero-parameter text and machine-readable entitlements. Real friction remains, and the positioning sidesteps the library channel rather than pretending it does not. |
| Unconditional use and fork freedom for any organisation, at any size | Intentionally different — that is the point. A card-trivial fee, a 60-day cure window, amnesty covenants on purchase; worst case, pin and wait four years for the Apache conversion. |
| Distribution and package-registry inclusion | A library concern. This licence is explicitly not for libraries, and four-year-old versions qualify normally. |
| Familiarity — zero marginal legal review | The review happens once per category, ever: one counsel-drafted text, zero variants, published memos. |
| Decades of litigated certainty | Three-jurisdiction memos precede launch, and payers are insulated by vesting regardless of any legal outcome. The uncertainty is real and it is ours, not the adopter’s. |
| No steward to fail or be captured | A lapse backstop, an independent-majority board, an append-only ledger, and a kill protocol that strands nobody. A steward is still a risk we added. |
| Some employers bar contributions to non-OSI licences | Real, not zero, and priced in — stated here rather than discovered by a contributor whose employer says no. |
| Dual licensing lets a maintainer monetise their own work | A deliberate boundary: revenue-seeking maintainers belong with BUSL or FSL. This converts a zero counterfactual into public benefit, never away from maintainer income. |
The five attacks this page will receive — answered here
Published pre-emptively, in full, on the page they are aimed at, rather than waiting to be asked. An answer written under pressure is worth less than one written in advance.
-
“Your comparison table implies parity. It is not the same thing.”
Correct, and the first content row of the comparison page says so before any tick mark appears. Comparison is not equivalence.
The table exists because the question maintainers actually ask is “what changes for my community?”, and the honest answer has three parts: a long list of things that do not change, a shorter list of things this adds, and a list of places where open source keeps a real edge. That third part is not collapsed, not hidden behind a toggle, and not placed below the fold — a build assertion fails the site if it ever is.
-
“Some employers bar contributions to licences like this, and Linux distributions will never ship it.”
Both true. Both priced in. Both stated here rather than discovered later.
Some open source programme offices bar contribution to any licence that is not OSI-approved, exactly as they do for every BUSL or FSL project. That is a real cost to a project that adopts this licence, and we will not claim it is zero — a single screenshot of an internal policy would refute us.
Two things soften it. The four-year Apache-2.0 conversion means even the strictest shop can consume, pin, and eventually contribute to older versions on ordinary permissive terms. And the licence follows function: this text is meant for deliberately-adopted products, not for libraries deep in a dependency tree, and our own ecosystem repositories stay MIT or Apache.
We would rather lose contributors honestly than win them by pretending the cost does not exist.
-
“Who audits you? How would anyone know the money actually reaches charity?”
The pledge has one wording and it is deliberately qualified: 100% of net proceeds after published, capped operating costs (cap: set with counsel before launch, audited). We never state it in an unqualified form, in any channel, because the unqualified form is false the moment a payment processor takes its cut.
Four mechanisms, none of which requires trusting us:
- The full fee stack is published as one figure — merchant-of-record fee, intermediary fee, capped operating levy, currency drag — before any journalist computes it for us.
- The allocation ledger is append-only. Corrections are new rows; annotations render as annotations. Monthly exports are immutable once written and carry a hash chain.
- Accounts are audited and the audit reports publish on the Trust Center as they exist.
- Neither the steward nor any repository owner can be a recipient. That is a structural bar in the statutes, not a policy we could quietly revise.
What we do not claim: that every step of the banking and intermediary leg is publicly provable end to end. It is not, and saying otherwise would be the easiest overclaim to puncture. The claim we do make is that every recorded allocation and disbursement is independently reconcilable — append-only ledger, audited accounts, partner receipts.
Until money has moved, there is nothing to audit and this site says exactly that.
-
“Every fee-for-commercial-use licence before this one died. Why would yours survive?”
It may not. The graveyard is our syllabus rather than our embarrassment: the post-mortem of prior attempts publishes with honest odds, not as marketing.
What is different is narrow and structural, and each item is checkable:
- One canonical text, zero parameters. The legal review happens once per category, not once per project. Every adopter ships the identical file.
- SPDX identifier requested at launch, so scanners can name the thing rather than flagging it as unknown.
- Machine-readable entitlements. A procurement team gets a signed record and a verification page, not an email thread.
- An escrow launch, so no first adopter stands alone: the licence goes live only when a published minimum of projects and payers is in place.
- Counsel memos in three jurisdictions before launch, published.
- The adopter’s worst case is capped in the text. Versions vest permanently, the four-year Apache-2.0 conversion is unconditional, and a steward-lapse backstop turns the licence permissive if this organisation stops operating.
We can fail. Adopters cannot be stranded. Those are two different risks, and only the first one is ours to take.
-
“For a large company the annual fee is a rounding error. Is this not just virtue-badge vending?”
The paperwork forbids it, which is the only answer worth giving.
The base credential is marketed as compliance, not generosity — the product is one flat annual licence, one review, machine-verifiable proof. Certificates ship with a claim-language kit that states the exact wording a buyer may use; generic “we support charity” framing is contractually excluded. Organisations holding a gratis waiver receive status certificates only, with no funding claim available to them at all.
Prominence is earned by real multipliers, not by the base fee: the voluntary impact-multiplier tiers exist precisely so that a company wanting a creditable story has to pay for one. We also never frame the fee as beneath a large organisation’s attention — scanners read a licence at zero cost, and a stealth framing would be both untrue and insulting.
Tiny cost per payer, distributed impact, and no permission to exaggerate either. That is the whole design.
Twenty more answers, grouped by who asks them: the FAQ. The same trade-offs written for one reader at a time: repository administrators · contributors · companies · the causes · the impact displays · the Association · news.