Open Source Infrastructure Deserves a Business Model  🔗 **Source:** [Open Source Infrastructure Deserves a Business Model](https://dri.es/open-source-infrastructure-deserves-a-business-model) > [!note]+ Document Notes > > [!info] Provenance > > **Analysis generated:** 2026-08-27 by Claude (Opus 5) via the `readwise-deep-read` pipeline. > > **Source:** dri.es — https://dri.es/open-source-infrastructure-deserves-a-business-model > > **Length:** 1,164 words · article · **Author:** **Dries Buytaert** (creator of Drupal, co-founder of Acquia) > > **Note on this document:** `[EVIDENCE]` Reader returned a **pre-existing doc ID** on save (`01kka7a…`), so this was saved some time ago and **never analysed** — notes were empty. It is the third of three linked posts developing one argument; the other two are analysed as `dries_open-source-cost-allocation-system.md` and `dries_license-only-versus-stewarded.md`. > > **Acknowledged reviewers:** `[STATED]` Tiffany Farriss, Tim Lehnen, Gábor Hojtsy, Lauri Timmanee. > > **Disclosed interest:** `[INFERENCE]` Acquia is the largest commercial beneficiary of Drupal; Buytaert also helps lead the project whose Association's $3M budget he cites. **The argument advocates a funding model his ecosystem would benefit from.** Noted for calibration, not as disqualification. > > **Register key:** `[STATED]` author-stated · `[EVIDENCE]` demonstrated · `[INFERENCE]` reviewer inference · `[RECOMMEND]` reviewer recommendation. > > > > ⚠️ **Contains a direct claim about FAIR.** See [The FAIR claim](#the-fair-claim-and-how-to-hold-it) — flagged prominently because it concerns a project you are involved in and I have no independent basis to assess it. > > # 🏷️ Title: Open Source infrastructure deserves a business model > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References Named in the piece (no outbound links captured): - **PyPI** / Python Software Foundation — donated infrastructure from Fastly, AWS, Google Cloud; PSF post *"last October"* on its fragility - **Packagist** / Private Packagist · **npm** / GitHub / Microsoft · **RubyGems** · **WordPress.org** / Automattic · **Drupal Association** · **GNOME** - **The FAIR project** — *"a federated package manager backed by the Linux Foundation"* ## ⚠️ Correction this forces on my earlier analysis `[EVIDENCE]` This morning, analysing the newest post in this series, I wrote: *"There is a real case against the framing and it appears nowhere: the cost diffusion is a feature, not a defect… He gets close and then treats the diffusion purely as a problem to be managed."* `[EVIDENCE]` **He engages it here, and rejects it directly:** > `[STATED]` *"In Open Source, what looks like fairness often is not. Free for everyone sounds equitable, but the cost does not disappear. It is absorbed by those who can least afford it, while the organizations that benefit most often pay the least. When a Fortune 500 company consumes Open Source infrastructure for free, that is not a neutral outcome. **It is a subsidy flowing in the wrong direction.**"* `[INFERENCE]` **My criticism was of an argument assessed in isolation that is actually one of three linked posts.** The counter-case I said was missing is answered in the piece the newer post links to — and answered well: the claim is not that diffusion is bad in the abstract, but that **"free for everyone" is regressive in practice**, because absorbing an unpriced cost is easiest for the parties with the most resources and hardest for the ones with the least. `[RECOMMEND]` **The general lesson is one this batch keeps producing:** a linked series is one argument, and assessing the newest instalment alone will find gaps that the earlier ones fill. **Read the "keep reading" links before criticising an omission.** ## 📝 Summary **The problem.** `[STATED]` *"Open Source software is free to download. But the infrastructure that makes it usable is not."* Registries, Git hosting, CI, release tooling — `[STATED]` *"Most of this infrastructure is invisible to end users, and almost no one thinks about what it costs to run."* **The survey is the article's substance**, and it is specific: | Project | Funding mechanism | Detail given | |---|---|---| | **PyPI** | Donated infrastructure | `[STATED]` *"billions of downloads a day"* on Fastly/AWS/Google Cloud; PSF says losing one sponsor would cost *"tens of thousands of dollars a month"* to replace | | **Packagist** | Single-company patronage | Cross-subsidised by the commercial Private Packagist | | **npm** | Single-company patronage | Independent → *"serious financial trouble"* → GitHub, 2020 → *"critical JavaScript infrastructure is now owned by Microsoft"* | | **WordPress** | Single-company patronage | `[STATED]` *"Automattic, by far the ecosystem's largest commercial beneficiary, subsidizes most of the infrastructure"* | | **RubyGems** | Multi-company sponsorship + community | `[STATED]` $2,500–$5,000/year, *"about 110 supporters needed"* | | **Drupal** | Donated + community + sponsorship | `[STATED]` *"roughly $3 million a year"* | **The observation that unifies them.** `[STATED]` *"none of these approaches tie funding directly to how much the infrastructure is used."* **The proposal.** `[STATED]` *"keep core infrastructure free for individuals and small projects, while organizations using it at scale help pay for what they consume. Not as a donation, but as payment."* **The thesis that matters most.** `[STATED]` *"governance tends to follow funding… When one organization pays for the infrastructure, it ultimately controls it. When a broader set of stakeholders funds it, governance broadens with it."* And the closing: `[STATED]` *"Solving the funding problem is a prerequisite to solving the governance problem."* **Bottom line.** `[INFERENCE]` The strongest of the three posts, because it carries data rather than only a frame. **Two things to hold carefully: the FAIR characterisation, and the fact that "governance follows funding" is both his central descriptive claim and the premise that justifies his proposal.** ## 📖 Main Content ### The FAIR claim, and how to hold it `[STATED]` *"The FAIR project, a federated package manager backed by the Linux Foundation, was designed to give the WordPress ecosystem an independent alternative. **The software works but its organizers recently stepped back after failing to secure long-term funding commitments.**"* `[STATED]` And, in the governance section: *"FAIR is a recent example. **The project didn't fail because federation was the wrong idea. It failed because nobody built a business case compelling enough to fund it as an alternative.**"* `[INFERENCE]` **Recorded verbatim because you have direct knowledge here and I have none.** What I can say about the claim as a claim: - **He is explicit that the technical work succeeded** — *"the software works"*, *"federation [was not] the wrong idea"*. The attribution of cause is entirely to funding, and he does not hedge it. - **It is doing argumentative work.** FAIR is the piece's clearest evidence for *"governance is downstream of funding"*, and it is the only cited case where a project attempting to change an ecosystem's infrastructure governance is said to have stopped for want of money. `[INFERENCE]` **A claim that is both load-bearing and about a third party deserves more sourcing than it gets** — no link, no statement, no named organiser. - **The word "failed" is his**, and it is stronger than *"stepped back"* two paragraphs earlier. `[INFERENCE]` The two framings are not the same and the piece uses them interchangeably. `[RECOMMEND]` **You are positioned to assess whether the funding account is complete, and I am not.** If it is incomplete — if governance resistance, timing, or scope also mattered — **that is worth saying publicly**, because this post is likely to become the canonical outside account of what happened, and it is written by the most prominent person in an adjacent ecosystem. ### "Governance follows funding" is the pivot, and it is convenient `[STATED]` *"Discussions about Open Source infrastructure often focus on governance: who should control it and who gets to make the decisions. In reality, those questions are often settled by something simpler: who pays for it."* `[INFERENCE]` **This is simultaneously his strongest descriptive claim and the premise that licenses his proposal**, which is a combination worth noticing. If governance follows funding, then broadening funding broadens governance — and his usage-fee proposal becomes a *governance* reform rather than a commercial one. **That is a lot of weight on a claim supported by one case.** `[EVIDENCE]` **His own table contains the counter-evidence.** RubyGems uses community fundraising across ~110 supporters; Drupal combines four mechanisms. `[INFERENCE]` Those are already broad-based, which means the relationship is **a spectrum rather than a rule**, and the interesting question — *how broad does funding have to be before governance actually broadens?* — is not asked. `[INFERENCE]` **And it sits in tension with the newest post in the series.** There, cost-bearers should have *"a way to make those costs visible and **influence** the decision, **even if they do not control it**."* Here: *"When one organization pays for the infrastructure, it ultimately controls it."* `[INFERENCE]` **The first restrains the funder; the second observes that funders control and uses it to argue for more funders.** They are compatible — one is normative, one descriptive — but **the restraint in the newer post is the more defensible position, and this piece does not carry it.** ### The WordPress row is the one that lands locally `[STATED]` *"WordPress.org runs on a different version of the same dynamic: corporate patronage. Automattic, by far the ecosystem's largest commercial beneficiary, subsidizes most of the infrastructure. **It works, but it also means that whoever funds the infrastructure controls it.**"* `[INFERENCE]` **Stated flatly, by an outsider, without polemic — and it is the structural explanation for a governance episode already in this batch.** The WPpodcast account of the 7.1 release day had the release lead publishing an announcement herself, with automatic updates already running across ~300,000 sites, after *"lack of communication from Mullenweg."* And The Repository reported the official WordPress account subsequently used the WP Rocket incident to promote Jetpack. `[INFERENCE]` **Dries is not commenting on either event and supplies the frame that makes them structural rather than personal:** if one company funds the infrastructure, control of it is not a governance question that remains open. `[RECOMMEND]` **That is the more useful way to hold those incidents** — not as decisions individuals made, but as what the funding arrangement predicts. ### GNOME is the sharpest example and gets three sentences `[STATED]` *"In February 2026, GNOME began redirecting Git traffic from its own GitLab to GitHub mirrors to reduce bandwidth costs. As a result, GitHub and its owner Microsoft now absorb some of GNOME's bandwidth cost."* `[INFERENCE]` **This is the whole argument compressed into one event, and it is the piece's most alarming data point.** A project that deliberately built independent infrastructure **gave part of it back to a platform because it could not afford the bandwidth.** Not a governance decision, not a strategic pivot — a cost decision with a governance consequence. `[INFERENCE]` And it composes with npm: *"critical JavaScript infrastructure is now owned by Microsoft."* **Two ecosystems, two routes, same destination.** The piece reports both and does not draw the line between them. ### The regressive-subsidy argument is the strongest paragraph `[STATED]` *"Free for everyone sounds equitable, but the cost does not disappear. It is absorbed by those who can least afford it, while the organizations that benefit most often pay the least… **It is a subsidy flowing in the wrong direction.**"* `[INFERENCE]` **This is the argument that makes the proposal defensible rather than merely commercial**, and it inverts the intuition it is arguing against. The objection to usage fees is that they make open source less free. The answer is that **"free" is already a price paid by someone, and the current allocation is regressive** — the unpaid labour and absorbed cost fall on maintainers and small foundations, while the largest consumers pay least. `[EVIDENCE]` He also pre-empts the reflexive objection with a historical parallel: `[STATED]` early fears that paid contributors would drive volunteers away, where *"in practice, the opposite happened."* `[INFERENCE]` And he immediately limits it — *"That does not mean every new funding idea is a good one. But instinctive discomfort alone is not a reason to reject it."* **Using a precedent to license consideration rather than to prove the case is the correct strength of that move.** ### What the proposal does not address `[EVIDENCE]` He is candid that the design is unspecified: `[STATED]` *"Exactly how that would work in practice is a separate design question… Questions about measurement, thresholds, and enforcement would need careful community discussion."* `[INFERENCE]` Three problems the piece names in passing and does not engage, each of which is where such schemes usually fail: - **Measurement.** Download counts are trivially inflatable and heavily automated — CI runs, mirrors, containers rebuilt per commit. `[INFERENCE]` **Billing on a number that is mostly machines re-fetching is a hard basis for an invoice.** - **Enforcement.** There is no mechanism to compel payment from someone the licence permits to redistribute freely. `[INFERENCE]` The scheme depends on voluntary compliance by exactly the parties currently not paying. - **The mirror escape.** `[INFERENCE]` **His own GNOME example is the counter-move**: an organisation facing usage fees can mirror the registry and pay nothing, which is what the licence explicitly allows. A usage fee on freely-redistributable software is a fee on *convenience*, not on the software — which may be enough, and is a much weaker lever than the framing implies. ## 🗒️ Notes - **The table is the most useful artifact in the three posts** — six ecosystems, four funding mechanisms, and the pattern visible at a glance. `[INFERENCE]` **Three of six are single-company patronage**, which is the finding the table makes without stating. - **The PyPI fragility figure is the most quotable:** `[STATED]` losing one sponsor would cost the PSF *"tens of thousands of dollars a month."* `[INFERENCE]` For infrastructure serving *"billions of downloads a day"*, that is a startlingly small number — and the vulnerability is not the amount but that it is **one renewal decision away.** - **RubyGems' numbers are the concrete alternative:** `[STATED]` $2,500–$5,000/year, ~110 supporters. `[INFERENCE]` That implies roughly **$275k–$550k/year** — an order of magnitude below Drupal's $3M, which suggests either very different scope or that the target is under-set. The piece does not say whether the programme reached its target. - **Drupal's $3M is the only full operating figure given**, and it covers *"Composer endpoints, GitLab repositories, CI pipelines, automatic update notifications, and more."* `[INFERENCE]` **Automatic update notifications** is the same service as the cost-allocation post's 50–75 TB example — **the three posts are describing one system from three angles.** - **The npm history is compressed to one sentence** and lands hard: independent company → financial trouble → acquisition → *"critical JavaScript infrastructure is now owned by Microsoft."* `[INFERENCE]` Stated as fact rather than complaint, which is more effective. - **The reviewer acknowledgements are worth noting** — `[STATED]` Tiffany Farriss, Tim Lehnen, Gábor Hojtsy, Lauri Timmanee. `[INFERENCE]` Drupal Association leadership and long-standing Drupal contributors, which means **the piece was reviewed inside the organisation whose $3M budget it cites** — appropriate for accuracy, and it does mean the critique is friendly-internal rather than independent. - **"Governance is downstream of funding"** is the section heading, and `[INFERENCE]` it is a stronger claim than the body supports — the body says governance *tends to* follow funding. ## 📚 References `[EVIDENCE]` **No outbound links captured**, despite citing a specific PSF blog post (*"last October"*), a dated GNOME infrastructure change (February 2026), the RubyGems funding programme, and the FAIR project. `[INFERENCE]` Every one of those is a checkable primary source and none is reachable from the piece. `[INFERENCE]` **For the FAIR claim in particular this matters**, because it is the load-bearing evidence for the governance thesis, concerns a third party, and is stated without attribution to any statement or announcement. `[RECOMMEND]` **Treat every figure here as accurate-until-checked rather than sourced** — the numbers are specific enough to be verifiable and none is verified in the text. ## 🗣️ Paraphrase Open source software is free to download; the infrastructure that makes it usable is not. Registries, Git hosting, CI and release tooling all cost money to run, and almost nobody thinks about it. Every large project has improvised something. PyPI handles billions of downloads a day on infrastructure donated by Fastly, AWS and Google Cloud — and the Python Software Foundation has said that if one sponsor declined to renew, replacing it would cost tens of thousands a month. Packagist is cross-subsidised by a commercial product from the company that runs it. npm tried to be independent, ran into financial trouble, and ended up owned by Microsoft. WordPress.org is largely subsidised by Automattic, the ecosystem's biggest commercial beneficiary. RubyGems asked businesses for $2,500–$5,000 a year and needs about 110 of them. Drupal's Association spends roughly $3 million a year. None of those ties funding to how much the infrastructure is actually used. When the economics break it shows: GNOME started redirecting Git traffic to GitHub mirrors in February 2026 to cut bandwidth costs, so Microsoft now absorbs part of it. His proposal is a value exchange — free for individuals and small projects, paid by organisations consuming at scale. Not donation, payment. He expects instinctive resistance and points out that people feared paid contributors would drive volunteers away, and the opposite happened. And he argues "free for everyone" is not actually fair: the cost is absorbed by whoever can least afford it while the biggest beneficiaries pay least — a subsidy flowing the wrong way. Finally, governance tends to follow funding. Whoever pays for the infrastructure controls it; broaden the funding and governance broadens too. He cites FAIR — the Linux Foundation-backed federated package manager for WordPress — as a case where the software worked but the organisers stepped back without long-term funding. Solving funding, he argues, comes before solving governance. ## 📰 Newsletter **The third of Dries Buytaert's linked posts is the one with data, and it contains a direct claim about FAIR.** > *"The FAIR project, a federated package manager backed by the Linux Foundation, was designed to give the WordPress ecosystem an independent alternative. The software works but its organizers recently stepped back after failing to secure long-term funding commitments."* And later: *"The project didn't fail because federation was the wrong idea. It failed because nobody built a business case compelling enough to fund it as an alternative."* **He attributes the cause entirely to funding and doesn't hedge it** — and it's the load-bearing evidence for his governance thesis, cited without a link or a statement. Anyone with direct knowledge is better placed than I am to say whether that account is complete, **and this post is likely to become the outside world's canonical version of what happened.** **The survey is the real substance.** PyPI runs billions of downloads a day on donated infrastructure, and the PSF says losing *one* sponsor would cost tens of thousands a month to replace. npm went independent → financial trouble → owned by Microsoft. **WordPress.org is subsidised by Automattic** — *"It works, but it also means that whoever funds the infrastructure controls it."* Drupal's Association spends **~$3M/year**. RubyGems asks businesses for $2,500–$5,000 and needs ~110 of them. **Three of his six examples are single-company patronage.** **The sharpest data point gets three sentences.** In February 2026 GNOME started redirecting Git traffic to GitHub mirrors to cut bandwidth costs — *"GitHub and its owner Microsoft now absorb some of GNOME's bandwidth cost."* **A project that built independent infrastructure gave part of it back to a platform because it couldn't afford the bandwidth.** Not a strategic pivot. A cost decision with a governance consequence. **And the strongest paragraph corrects something I said this morning.** I criticised the newest post for never engaging the case that cost diffusion is a feature. **He engages it here:** *"Free for everyone sounds equitable, but the cost does not disappear. It is absorbed by those who can least afford it, while the organizations that benefit most often pay the least… It is a subsidy flowing in the wrong direction."* **I assessed one instalment of a three-part argument in isolation.** Read the "keep reading" links before calling something an omission. **One thing to push on.** *"Governance follows funding"* is both his central descriptive claim and the premise that licenses his proposal — and his own table has counter-evidence: RubyGems and Drupal already fund broadly. It's a spectrum, and the interesting question — **how broad does funding have to be before governance actually broadens?** — isn't asked. It also sits oddly against the newest post, where cost-bearers get *"influence… even if they do not control it."* **That restraint is the better position, and this piece doesn't carry it.** ## ✅ Action Items #### DevRel - **Developer advocacy:** `[RECOMMEND]` The **GNOME example** is the most compressed argument available for why infrastructure funding is a governance issue — one project, one bandwidth decision, one loss of independence. - **Education:** `[RECOMMEND]` Teach the **regressive-subsidy** framing. It inverts the intuition that "free" is neutral, and it does so with a mechanism rather than an appeal. - **Developer marketing / Developer experience / Community management / Cross-project:** Not applicable to this source. #### OSPO - **Policy & Governance:** `[RECOMMEND]` **Ask which funding mechanism your ecosystem's infrastructure actually uses**, using his four categories. `[INFERENCE]` If the answer is single-company patronage, the governance question is already settled whatever the charter says. - **Strategic Alignment:** `[INFERENCE]` The uncomfortable corollary for any organisation depending on someone else's registry: **your dependency's independence is only as durable as its funding.** - **Software Inventory:** `[RECOMMEND]` Infrastructure operating cost belongs in inventory. `[EVIDENCE]` Drupal's $3M is the only full figure in the piece, and it is the kind of number that makes the argument concrete. - **Community Engagement:** `[INFERENCE]` The RubyGems model — $2,500–$5,000/year, ~110 supporters — is the most replicable of the four, and the piece does not say whether it succeeded. **Worth finding out before copying it.** - **License Compliance / Security Management / Developer Support:** Not applicable to this source. #### Community - **Sponsors:** `[RECOMMEND]` The subsidy-direction argument is the strongest available case for corporate funding: **not charity, correction of a regressive allocation.** - **Maintainers:** `[INFERENCE]` The mirror escape means a usage fee is a fee on **convenience**, not on software. Anyone designing one should assume the largest consumers can route around it — which is what the licence permits. - **Ecosystem:** `[INFERENCE]` **Three of six examples are single-company patronage**, and two of those companies are Microsoft. That concentration is the table's unstated finding. - **Contributors:** Not applicable to this source. #### WordPress - **Meta:** `[STATED]` *"Automattic… subsidizes most of the infrastructure. It works, but it also means that whoever funds the infrastructure controls it."* `[INFERENCE]` **This is the structural explanation for the governance episodes already in this batch** — the 7.1 release-day decision log and the Jetpack promotion during the WP Rocket incident. Not personal decisions; what the funding arrangement predicts. - **Hosting:** `[INFERENCE]` Hosts are among the largest consumers of api.wordpress.org and contribute nothing to its operation — the clearest instance of the regressive-subsidy claim inside WordPress. - **Core:** `[INFERENCE]` WordPress's batched update-check architecture is a cost-allocation advantage over Drupal's N+1 (verified in source earlier today). **Worth knowing that the ecosystem's infrastructure bill is lower than it could be**, and that this was never framed as a funding decision. - **All other WordPress teams:** Not applicable to this source. #### The WPOCC - **Funding experiments:** `[RECOMMEND]` **The most directly relevant source in the batch.** Four named mechanisms with real numbers, and the observation that unites them — *"none of these approaches tie funding directly to how much the infrastructure is used."* `[INFERENCE]` **The design space is genuinely open**, and RubyGems' figures are the concrete comparator. - **Programs:** `[INFERENCE]` *"Solving the funding problem is a prerequisite to solving the governance problem"* is a sequencing claim worth testing rather than assuming — it implies governance work without funding work is premature, which is a strong position with obvious counter-examples. - **Projects:** Not applicable to this source. ### FAIR `[RECOMMEND]` **Read the two quoted passages directly** — they are recorded verbatim under [The FAIR claim](#the-fair-claim-and-how-to-hold-it). - `[STATED]` The characterisation is that **the software worked and the funding did not** — *"didn't fail because federation was the wrong idea."* - `[INFERENCE]` **He is generous about the technical work and unequivocal about the cause.** Whether the cause is complete is a question you can answer and I cannot. - `[INFERENCE]` **It is cited without a link, a statement, or a named source**, and it is the piece's single strongest piece of evidence for the governance thesis. `[RECOMMEND]` Given that this is likely to become the canonical outside account, **a public correction or amplification from someone with direct knowledge would be load-bearing** — in either direction. - `[INFERENCE]` **The structural argument, independent of the FAIR specifics, is the useful part:** *"When one organization pays for the infrastructure, it ultimately controls it."* That is the clearest statement in this batch of what an independent distribution layer would actually have to solve — **not the technology, the funding**, and the technology is reported here as having worked. ## 🧪 Hypotheses 1. **H1 — Governance breadth tracks funding breadth.** Across ecosystems, the number of independent funders of core infrastructure predicts the distribution of decision authority over it. 2. **H2 — Single-company patronage is the modal arrangement.** A majority of major package registries and ecosystem infrastructure depend primarily on one company. 3. **H3 — Usage fees are avoidable by mirroring.** Any usage-based fee on freely-redistributable software is routed around by the largest consumers within a short period, because the licence permits it. 4. **H4 — Download counts are a poor billing basis.** A large majority of registry downloads are automated (CI, containers, mirrors) rather than distinct human-initiated installs. 5. **H5 — The subsidy is regressive.** Infrastructure cost borne per unit of consumption is inversely related to the consumer's resources. 6. **H6 — Funding precedes governance change.** Attempts to change ecosystem infrastructure governance without independent funding fail at a higher rate than funded attempts. ## 📊 Methodology - **H1 — the claim the whole argument rests on, and it is testable.** Code a sample of ecosystems on funder count and on distribution of decision rights (charter, board composition, commit authority); test the correlation. **Confound:** both correlate with project age and size, so those must be covariates or you measure maturity. - **H2 — nearly done already.** His table is six cases; extend it to twenty registries with sourced funding data. `[RECOMMEND]` **Cheap, mechanical, and it would turn the table's implicit finding into a citable one.** - **H3:** Observe behaviour if any registry implements usage fees. `[INFERENCE]` **No current instance to observe** — which is itself the finding, and makes this prospective rather than testable now. - **H4 — immediately runnable and it undercuts the proposal's mechanism.** Registries publish download data; classify by user agent and request pattern. `[INFERENCE]` **If most downloads are machines, usage-based billing is measuring the wrong thing**, and this is the cheapest way to find out. - **H5 — the empirical version of the article's strongest paragraph.** Requires per-consumer usage and per-consumer contribution data, which registries hold and rarely publish. **Confound:** large consumers often contribute in non-financial ways (code, staff time), so "pay least" needs a definition covering contribution in kind. - **H6:** Case comparison of infrastructure-governance initiatives by funding status. **Weak design, small n** — and `[INFERENCE]` **FAIR is currently n=1 for the unfunded arm**, which is exactly why the piece leans on it so heavily. ## ❓ Questions - **Is the FAIR account complete?** `[INFERENCE]` The single most consequential question in this piece, and one you can answer. - **Did RubyGems' programme reach its ~110 supporters?** `[EVIDENCE]` The target is given; the outcome is not — and the outcome is what makes the model replicable or cautionary. - `[INFERENCE]` **What stops a large consumer from mirroring?** The licence permits it and his own GNOME example demonstrates the move. - What fraction of registry downloads are automated? `[INFERENCE]` If most are, usage-based fees bill for machine traffic. - `[INFERENCE]` **How broad must funding be before governance broadens?** He asserts the relationship and never characterises its shape. - Does the newest post's *"influence, not control"* restraint survive this piece's *"whoever funds controls"*? `[INFERENCE]` They are compatible and the emphasis differs, and he never reconciles them. ## 🔮 Future Research - **Comparative infrastructure economics with sourced numbers** across twenty-plus ecosystems. `[INFERENCE]` His table is the best existing version and it is six rows and unsourced. - **Whether usage-based funding is defeatable by mirroring.** `[INFERENCE]` The design question his proposal defers, and the one that decides whether it can work at all. - **Governance breadth as a function of funder count.** The thesis, currently supported by one case. - **The composition of registry download traffic.** `[INFERENCE]` Prerequisite for any usage-based model, and publicly answerable. ## 💡 Implications - **Practice.** `[RECOMMEND]` Identify which of the four mechanisms funds the infrastructure you depend on. `[INFERENCE]` **If it is single-company patronage, plan for the possibility that the company's interests and yours diverge** — GNOME and npm are the two worked examples. - **Policy.** `[INFERENCE]` **"Free for everyone" is a funding decision, not the absence of one**, and its incidence is regressive. That reframing is available to any project without adopting his proposal. - **Security.** `[INFERENCE]` Underfunded infrastructure is a supply-chain risk. `[EVIDENCE]` PyPI is one non-renewal from a five-figure monthly bill, and it serves billions of downloads a day. - **Education.** `[EVIDENCE]` The paid-contributor precedent is used correctly — **to license consideration, not to prove the case** — and explicitly limited: *"instinctive discomfort alone is not a reason to reject it."* - **Funding models.** `[INFERENCE]` **The sequencing claim is the sharpest thing here and the least examined:** *"Solving the funding problem is a prerequisite to solving the governance problem."* If true, governance work without funding work is premature — which is a strong claim with obvious counter-examples, and it is stated as a conclusion rather than argued. ## 🧠 Learning Styles Assessment **Reading level estimate.** `[INFERENCE]` **Flesch–Kincaid grade 11–12** — plain vocabulary, moderate sentences, low subordination. Hand estimate on 1,164 words, ±1.5 grades. `[EVIDENCE]` Registry names (npm, Composer, pip, Cargo, Packagist) are used without gloss, which assumes a developer reader — a gentler assumption than the surrounding argument requires, since the funding argument is legible to a non-developer and the examples are not. #### 👀 Visual learners - **Present:** `[EVIDENCE]` **A real comparison table** — six projects × four funding mechanisms — which is the best visual artifact in the three-post series, plus a metaphorical image (a cracking circular machine) with descriptive alt text. - **Improve:** `[INFERENCE]` The **dollar figures are scattered through prose** ($3M, tens of thousands/month, $2,500–$5,000 × 110) and never brought into the table, where they would make the scale differences immediate. #### 🔊 Aural learners - **Present:** Nothing. - **Improve:** `[INFERENCE]` Low value for a table-centred argument; the table does not survive audio at all. #### 🗣️ Verbal learners - **Present:** `[EVIDENCE]` **Strong.** *"It is a subsidy flowing in the wrong direction."* *"Governance is downstream of funding."* *"Infrastructure that entire ecosystems depend on cannot rely indefinitely on goodwill alone."* - **Improve:** `[INFERENCE]` *"Governance is downstream of funding"* as a heading overstates the body's *"tends to follow"* — the memorable version is the stronger claim, which is how the overstatement will propagate. #### 🧩 Kinesthetic learners - **Present:** `[EVIDENCE]` The four-category framework is directly applicable to any ecosystem a reader cares about. - **Improve:** `[EVIDENCE]` **The proposal is explicitly deferred** — *"exactly how that would work in practice is a separate design question."* `[INFERENCE]` A reader persuaded by the argument has nothing to implement. #### 🧠 Logical learners - **Present:** `[EVIDENCE]` Problem → survey → pattern → proposal → objection → governance implication, with the survey doing genuine work rather than illustrating a foregone conclusion. - **Improve:** `[EVIDENCE]` **The central thesis has one supporting case (FAIR) and counter-evidence in his own table** (RubyGems, Drupal already fund broadly). `[INFERENCE]` And the mirror escape is demonstrated by his own GNOME example without being connected to the proposal it undermines. #### 🧑‍🤝‍🧑 Social learners - **Present:** `[EVIDENCE]` Four named reviewers acknowledged, and six ecosystems' arrangements described with evident familiarity. - **Improve:** `[EVIDENCE]` **Nobody from any of those projects is quoted**, including FAIR — whose organisers are characterised as having *"stepped back"* without appearing. #### 🧘‍♀️ Solitary learners - **Present:** Self-contained; the table can be applied without further reading. - **Improve:** `[EVIDENCE]` **No links to any cited source** — the PSF post, the GNOME change, the RubyGems programme, FAIR. `[INFERENCE]` Every checkable claim requires the reader to go find it. #### 🌿 Naturalistic learners - **Present:** `[EVIDENCE]` The image — a machine cracking under pressure — and *"fragile arrangements"*, *"patchwork"*. - **Improve:** `[INFERENCE]` **The commons framing is the obvious fit and is still unused across all three posts.** Diffuse costs, concentrated benefit, no mechanism to price the externality — the vocabulary and the literature both exist, and *"subsidy flowing in the wrong direction"* is a restatement of it. ## 📊 Accessibility Assessment Assessed against **WCAG 2.2 AA**, from the captured markup only. `[INFERENCE]` **Scope limit:** content structure and text as retrieved. Focus, keyboard operation, live contrast, and reflow require the rendered page and are **not assessed**; absence of a finding is not a pass. #### Screen reader semantics - **Present:** `[EVIDENCE]` **Two `<h2>` headings in the body** — *"A patchwork of fragile arrangements"* and *"From steward to service provider"*, plus *"Governance is downstream of funding"*. `[INFERENCE]` **This one passes 2.4.6**, unlike the newest post in the series, which had headings only for "Keep reading". Descriptive and well placed. - **Present:** `[EVIDENCE]` Descriptive alt text on the lead image: *"A person stands before a massive circular machine with cracks forming inside it, suggesting infrastructure under pressure."* `[INFERENCE]` Conveys both content and intent — consistently good practice across this author's posts. - **Improve:** `[EVIDENCE]` **The funding table renders in the extraction as an undifferentiated run of cells** — headers and checkbox glyphs (☑/☐) flowing together with no visible row structure. `[INFERENCE]` If the underlying markup is a real `<table>` with header cells this is an extraction artifact; **if the checkmarks are unlabelled glyphs, each cell's meaning depends on visual position** and the table is not readable in sequence. **WCAG 1.3.1 Info and Relationships (A)** — flagged as needing checking against the rendered page rather than asserted. #### Keyboard navigation - **Not testable from the captured markup.** #### Contrast and readability - **Not testable** — no CSS captured. #### Text resizing and reflow - **Not testable from the captured markup.** `[EVIDENCE]` A six-column table is the standard 1.4.10 hazard and must scroll within its own container at 320px. Flagged as the thing to check. #### Media and captions - **Not applicable** — no audio or video. #### Forms and interactions - **Not applicable.** #### Standards reference - **1.1.1 Non-text Content (A)** — **satisfied**, and well. - **2.4.6 Headings and Labels (AA)** — **satisfied.** Three descriptive body headings. - **1.3.1 Info and Relationships (A)** — **indeterminate**; depends on whether the table is marked up as one and whether the ☑/☐ glyphs carry accessible names. - **1.4.10 Reflow (AA)** — **not assessable**; the table is the risk. ## 🌍 DEIB Assessment #### Cultural breadth - **Present:** `[EVIDENCE]` Six ecosystems spanning multiple languages and governance cultures. - **Improve:** `[INFERENCE]` All six are Western-led with US or European corporate funding, and the proposal — organisations paying for consumption — assumes consumers who can be billed in a currency and jurisdiction. `[INFERENCE]` **The "free for individuals and small projects" carve-out is where everyone else lands**, and how that line gets drawn across economies is exactly the design question he defers. #### Inclusive language - **Present:** Neutral and precise; no party is disparaged, including Microsoft, which appears twice as the beneficiary of others' cost problems. - **Improve:** Nothing substantive. #### Neurodiversity support - **Present:** `[EVIDENCE]` **Better structured than the other two posts** — three descriptive headings, a table, short paragraphs, and a clear linear argument. - **Improve:** `[INFERENCE]` The dollar figures are scattered rather than tabulated, so building the comparison requires holding four numbers across several sections. #### Representation - **Present:** `[EVIDENCE]` Four reviewers named and thanked. - **Improve:** `[EVIDENCE]` **Every named individual is a reviewer of the draft; no one from the six ecosystems speaks.** `[INFERENCE]` Most consequentially, **FAIR's organisers are characterised without being quoted or cited** — the piece describes their project as having failed and its cause, entirely in the author's voice. #### Psychological safety - **Present:** `[EVIDENCE]` **Handled carefully where it matters.** The FAIR passage is explicit that the technical work succeeded — *"the software works"*, *"didn't fail because federation was the wrong idea"* — which separates the people from the outcome. `[INFERENCE]` Nobody is blamed anywhere in the piece, including the sponsors who might not renew. - **Improve:** `[INFERENCE]` **Being named as the example of a failed project is a cost borne by people who did not choose to be the illustration.** `[INFERENCE]` The characterisation is generous, and it is still an outside account of someone else's project stated as settled fact, unsourced, by a prominent figure — **which is a different thing from criticism and can be harder to correct.** ## ✅ Logical Validity #### Strong reasoning - **The survey earns the conclusion.** Six ecosystems, four mechanisms, and the pattern — *"none of these approaches tie funding directly to how much the infrastructure is used"* — is derived from the data rather than asserted ahead of it. - **The regressive-subsidy argument inverts its opponent's premise rather than dismissing it.** *"Free for everyone sounds equitable, but the cost does not disappear."* `[INFERENCE]` It grants the intuition and shows the incidence, which is the strongest available form of that argument. - **The paid-contributor precedent is correctly limited.** `[STATED]` *"That does not mean every new funding idea is a good one. But instinctive discomfort alone is not a reason to reject it."* `[INFERENCE]` Using history to license consideration rather than to settle the question. - **GNOME and npm are well-chosen** because they are **outcomes, not opinions** — two projects that lost infrastructure independence for financial reasons. #### Fallacies - **Load-bearing claim on a single unsourced case.** FAIR carries the governance thesis, with no link, statement, or named source. **Weakness:** the strongest evidence is the least verifiable, and it concerns a third party. **Fix:** cite the announcement. - **Counter-evidence present and unaddressed.** RubyGems and Drupal already fund broadly, which makes *"whoever funds controls"* a spectrum rather than a rule. **Fix:** characterise the relationship's shape rather than asserting it. - **Self-undermining example.** `[INFERENCE]` GNOME demonstrates the **mirror escape** — an organisation reducing its costs by shifting traffic elsewhere — which is precisely the move a usage fee invites, and the piece never connects the two. - **Heading overstates body.** *"Governance is downstream of funding"* against *"governance tends to follow funding."* `[INFERENCE]` The memorable version is the stronger one. - **Terminology drift.** *"Stepped back"* (paragraph one) becomes *"failed"* (governance section) for the same event. **Weakness:** these are not the same claim, and the stronger one appears where it does more argumentative work. #### Tone of voice `[EVIDENCE]` Measured, data-forward, and unusually free of advocacy heat for a piece proposing that people start paying for something currently free. `[INFERENCE]` The restraint is strategic and effective: **the survey does the persuading**, and by the time the proposal arrives the reader has already accepted the problem. `[EVIDENCE]` Confidence is mostly calibrated — *"One direction that makes sense to me"*, *"one possible approach"*, *"would need careful community discussion"* attach to the proposal. `[INFERENCE]` **The exception is the FAIR claim**, which is stated with no hedge at all, and it is the one claim about a third party's internal situation. **The hedging pattern runs opposite to where the evidence is weakest.** ## 🛡️ DARVO Analysis **No DARVO pattern present.** `[EVIDENCE]` DARVO — Deny, Attack, Reverse Victim and Offender — describes a response to accusation. This is an analytical argument with no accusation and no accused party. Checking each element explicitly rather than omitting the section: - **Deny:** None. `[EVIDENCE]` The piece affirms an unfunded-infrastructure problem in the author's own ecosystem and supplies its budget figure. - **Attack:** None. `[EVIDENCE]` **Conspicuously restrained where attack was available.** Microsoft appears twice as the beneficiary of other projects' cost failures (npm, GNOME) and is described neutrally both times. Automattic's control of WordPress infrastructure is stated as structural — *"It works, but…"* — rather than criticised. Sponsors who might not renew are not blamed. - **Reverse victim/offender:** None. `[EVIDENCE]` Maintainers and foundations absorbing unpriced costs are described as bearing a **misallocation**, not as wronged parties, and no counterparty is cast as an offender. `[INFERENCE]` Two adjacent observations, recorded separately because neither is DARVO: 1. **The argument advantages the author's position.** A usage-fee model benefits well-resourced ecosystems and the vendors within them, and Acquia is the largest commercial beneficiary of Drupal. **The affiliation is public and the reasoning stands independently**, but the alignment is worth holding. 2. **FAIR is characterised without representation.** `[INFERENCE]` Describing another project's outcome and its cause, unsourced, in your own voice is not an attack — the framing is generous about the technical work — **but the people described have no presence in the account.** Already logged under Representation. ## 🎯 Conclusion **Bottom line.** `[INFERENCE]` The strongest of the three linked posts, because it carries data rather than only a frame — and the one with a direct stake for you. 1. **The FAIR claim, recorded verbatim because you can assess it and I cannot.** `[STATED]` *"The software works but its organizers recently stepped back after failing to secure long-term funding commitments."* And: *"The project didn't fail because federation was the wrong idea. It failed because nobody built a business case compelling enough to fund it as an alternative."* `[INFERENCE]` **He is generous about the technical work and unequivocal about the cause, with no link, statement, or named source** — and it is the single strongest piece of evidence for his governance thesis. `[RECOMMEND]` **This is likely to become the canonical outside account.** If the funding story is incomplete, a public correction from someone with direct knowledge would be load-bearing. 2. **The survey is the substance, and its unstated finding is concentration.** PyPI is one non-renewal from a five-figure monthly bill while serving billions of downloads a day. npm went independent → financial trouble → **owned by Microsoft**. GNOME redirected Git traffic to GitHub mirrors in February 2026 to cut bandwidth — **so Microsoft absorbs part of it now.** `[INFERENCE]` **Three of six examples are single-company patronage, and two of those companies are Microsoft.** 3. **The WordPress row is the local one.** `[STATED]` *"Automattic… subsidizes most of the infrastructure. It works, but it also means that whoever funds the infrastructure controls it."* `[INFERENCE]` Stated flatly by an outsider — and **it is the structural explanation for the governance episodes already in this batch**, the 7.1 release-day decision log and the Jetpack promotion during the WP Rocket incident. Not individual decisions; what the funding arrangement predicts. 4. **And it corrects me.** `[EVIDENCE]` This morning I criticised the newest post for never engaging the case that cost diffusion is a feature. **He engages it here** — *"Free for everyone sounds equitable… It is a subsidy flowing in the wrong direction."* `[INFERENCE]` **I assessed one instalment of a three-part argument in isolation.** The rule that falls out: **read the "keep reading" links before calling something an omission.** **Two things to push on.** `[INFERENCE]` *"Governance follows funding"* is both the descriptive claim and the premise licensing the proposal, supported by **one case**, with **counter-evidence in his own table** — RubyGems and Drupal already fund broadly, so it is a spectrum and the shape is never characterised. And **his own GNOME example demonstrates the mirror escape**: a usage fee on freely-redistributable software is a fee on *convenience*, which the licence permits anyone to route around. The piece never connects the two. **Practical next steps.** 1. **Read the two FAIR passages directly** and decide whether the funding account is complete. 2. `[RECOMMEND]` Apply his four-category framework to WordPress infrastructure explicitly. `[INFERENCE]` If the answer is single-company patronage, **the governance question is already settled whatever any charter says** — which is the most useful thing this piece offers. 3. Check whether RubyGems' ~110-supporter programme actually succeeded before treating it as the replicable model. 4. `[INFERENCE]` Hold the sequencing claim — *"solving the funding problem is a prerequisite to solving the governance problem"* — as a **hypothesis worth testing**, not a conclusion. It implies governance work without funding work is premature, which is strong and stated rather than argued. <!-- deep-read:sections end --> 📄 **Full text:** ![[Full Document Contents/Articles/Open Source Infrastructure Deserves a Business Model|Open Source Infrastructure Deserves a Business Model, Dries Buytaert]] ## Highlights - Open Source infrastructure deserves a business model ([View Highlight](https://read.readwise.io/read/01kkaeh8tv4ktep532j9yk18k0)) - if a single sponsor decides not to renew, it would cost them tens of thousands of dollars a month to replace the lost infrastructure. ([View Highlight](https://read.readwise.io/read/01kkaeaz377fjnasb2zm0c17h7)) - Revenue from the paid product subsidizes the free public registry. It's one of the more sustainable models out there, though it means a public good depends on one company's continued success. ([View Highlight](https://read.readwise.io/read/01kkaec0b8q2qx9r6n8rdb0gxv)) - The end result is that critical JavaScript infrastructure is now owned by Microsoft. ([View Highlight](https://read.readwise.io/read/01kkaebgebedb95cbg05hykhhf)) - it also means that whoever funds the infrastructure controls it. ([View Highlight](https://read.readwise.io/read/01kkaeccwykmzrn7xzdncec184)) - The [FAIR project](https://fair.pm/), a federated package manager backed by the Linux Foundation, was designed to give the WordPress ecosystem an independent alternative. The software works but its organizers recently stepped back after [failing to secure long-term funding commitments](https://joost.blog/fair-wordpress-and-knowing-when-to-stop/). ([View Highlight](https://read.readwise.io/read/01kkaectvkn846158btc9epss2)) - FAIR is a recent example. The project didn't fail because federation was the wrong idea. It failed because nobody built a business case compelling enough to fund it as an alternative. ([View Highlight](https://read.readwise.io/read/01kkaeej55bg8smzfadb9s8yea))