A persistent and rhetorically seductive claim holds that the State of Texas is, in a meaningful sense, an open-source software project — that its statutes are “source code,” its legislature a “merge pipeline,” its courts a “revert of bad commits,” its constitution a “bootloader,” and its citizens “maintainers.” This essay refutes that claim in its entirety, and in doing so resolves the question to a single mechanical discriminator.
The claim fails on every axis that can carry it, and it fails in a way that is fully specified rather than merely asserted. The decisive fact is structural and checkable: the State of Texas has no statewide initiative for constitutional amendment or statute. The only path to the state’s supreme legal file runs through a two-thirds vote of both houses of the incumbent legislature — an un-waivable gate. Every route by which Texas law actually changes is staffed by the incumbent’s own organs; every power held by the governed is a veto, a ratification, or a selection from a menu the incumbent defines. That is not the architecture of open source. It is the architecture of source-available software: a maximally permissive license axis wrapped around a fork axis that is closed to the governed by an incumbent gate no party subject to the system can bypass.
The correct verdict is therefore source-available, not open source. The text of the law is public, freely redistributable, and modifiable in principle. The enforceable fork is not. This essay establishes the operative definition, then refutes every possible line of defense for the opposite conclusion — the readability argument, the interpreter argument, the maintainer argument, the amendment argument, the empirical-fork argument, the transactional argument, the judicial-appointment argument, the coercion argument, the definitional argument, the historical argument, and the license-axis argument — before specifying the complete mechanical account of the verdict.
1. What “Open Source” Actually Requires: The Fork Test
Any serious use of the term “open source” must be able to do the work the term exists to do: it must discriminate. A definition that applies to every constitutional republic, or to every jurisdiction with a published statute book, is not a definition; it is a synonym for “has a legal code,” and it concedes the question by refusing to ask it. The burden on any proponent of the thesis that “Texas is open source” is therefore to supply a definition that (a) includes Texas and (b) excludes the states they would not wish to call open source — state-socialist and authoritarian jurisdictions included. This burden is the load-bearing wall of the entire argument, and it is the thing the argument never successfully crosses.
1.1 The Operative Definition: Gate-Bypassability
The Open Source Definition (OSD), maintained by the Open Source Initiative, enumerates ten criteria governing redistribution, source distribution, derived works, integrity of author’s source, nondiscrimination, license transmission, and technology neutrality. A common error — committed by nearly every defender of the “Texas is open source” thesis — is to read these ten criteria as ten independent clauses, and to conclude that a legal text that is freely redistributable and modifiable in principle “clears” the definition. This reading mistakes the anatomy for the function.
The ten criteria are not ten separate promises. They are ten ways of guaranteeing one downstream fact: that a competent third party can take the source, build a modified copy, and run it, independently of the incumbent maintainers’ consent. The entire point of granting free redistribution, of requiring source in preferred form, of permitting derived works, is that these permissions aggregate into a capability the definition exists to secure — the capability to fork, build, and run. The license is the means; the fork is the end; and the OSD judges the means by the end it serves.
This yields the operative test, precise where the slogan is vague:
A system is open source if and only if the parties who operate under its governing source can diverge that source and run a modified copy without the permission of the incumbent operators.
I call this the fork test, and its discriminating quantity is gate-bypassability: not whether a gatekept merge process exists (it exists in every system, open source included), but whether the gate can be bypassed by unilateral action. In open source, a rejected pull request is a formality — the would-be contributor forks the repository and ships a divergent copy, needing no one’s consent. The gate exists, but it is elastic: it can be stepped around. In a system that is only source-available, the gate is a wall: there is no unilateral path around it, because there is no permitted divergence that does not first negotiate the incumbent’s approbation.
The fork test, so operationalized, does the discriminating work the term requires. It excludes the authoritarian jurisdictions whose statute books are public but whose governing source cannot be diverged by the governed (North Korea, the People’s Republic of China, the Kingdom of Saudi Arabia all fail). It excludes every jurisdiction whose top-level legal file is sealed behind an incumbent authoring gate. And — as section 5 will establish — it excludes Texas, on a fact that is checkable and decisive.
1.2 Why the License Axis Cannot Substitute for the Fork Axis
The tactical retreat of the weaker defense is to abandon the fork axis and declare victory on the license axis alone: “the law is public domain, so it is the most permissive license that exists, more permissive than MIT, and therefore open.” This move fails on two grounds, one logical and one architectural.
The logical ground is coupling. The OSD’s criteria are not severable; they are collectively the definition of a forkable artifact. To cleave the license axis from the fork axis is to define “open” as “permissible to copy,” and then to ignore that the permission is a promise the system is architecturally unable to honor. A GPL license attached to a program whose sole build machine compiles for the incumbent alone, and whose only running instance is operated against you, does not make that program open source. It grants a permission the machine refuses. The license says “you may”; the architecture answers “there is no compile target for you.” A license whose permissions the system cannot discharge is a license in name only — it is the available half of “source-available,” and it is precisely the half that was never in dispute.
The architectural ground is that even the license axis, taken purely on its own terms, guarantees nothing about openness of outcome. Public domain — the actual status of uncopyrightable statutory text — is the license that sanctions proprietary forks. It imposes no copyleft obligation; it permits anyone to take the source private. A legal regime is thus consistent, on the license axis alone, with an entirely closed enforcement apparatus. The most permissive license axis in existence is compatible with the least open fork axis that exists. That is not evidence for the thesis; it is the exact distribution the thesis’s defeat predicts.
The license axis discriminates nothing — every statute book on Earth clears it, because it tests only whether text can be copied and modified, not whether the governing meaning of the system can be diverged and run. The fork axis discriminates everything — it tests whether the governed can bypass the incumbent’s gate, and it is the axis on which Texas fails.
2. The Readability Fallacy: Publication Is Necessary, Not Sufficient
The most primitive defense of the thesis is the readability claim: the laws are online, searchable, and complete, therefore “the source is open.” This is true and it is worthless.
Publication is a necessary condition of openness, not a sufficient one. Every state publishes its laws. North Korea publishes its laws. Saudi Arabia publishes a code. The People’s Republic of China publishes a statute book, and its administrative regulations are publicly promulgated. If “publishes its governing text” were the test, then open source would be coextensive with the existence of written law, the term would denote nothing, and the question would be answered by tautology. The readability claim cannot be the discriminator because it does not discriminate.
The readability defense also commits the error of conflating observation with access. Reading the output of a system — even reading it completely, even reading “the program in the act of running” — is not the same as obtaining its source, and obtaining its source is not the same as being able to run a modified copy. If I watch a closed binary execute, I have acquired observation, not source, and, having acquired the source, I have acquired neither a compile target nor the right to diverge the running instance. “Observation is access to source” is asserted, never demonstrated, under this defense; and it is false. Access to source is the precondition of forking; observation is not even a step toward it.
2.1 The Hidden Build: The Discretionary Enforcement Layer
The readability defense’s strongest form concedes that the text is public and denies that anything material is hidden: “there is no private Texas; nobody holds a secret edition.” This is empirically false. The state’s execution of its source — the actual determination of whether and how the code runs against a given subject — is conducted in a layer that is substantially undisclosed:
- The prosecutor’s decision to charge or decline is discretionary, largely unpublished, and effectively unreviewable.
- Charging and plea bargaining occur in chambers and produce binding outcomes without public reasoning; the overwhelming majority of criminal cases in Texas resolve by plea, not trial, and the terms are negotiated where the record does not reach.
- Jury deliberations are sealed by law and by design — a black box at the very center of the “test suite” the metaphor invokes.
- Officer discretion — to arrest, to escalate, to use force — is exercised in the field and recorded, at best, after the fact.
The “no hidden Texas” claim is therefore false at its point of maximum confidence. The text is public; the discretionary interpretation that turns text into detention — the sovereign act that decides whether the code runs on you — is the hidden build. This is not a footnote; it is the operative layer. A system whose governing text is fully public but whose operative execution is sealed against the governed is not open source; it is precisely the “black box with a published specification” that the readability defense claims to have abolished. It has moved the box. It has not opened it.
3. The Interpreter Fallacy: Law Is Not Self-Executing
A more sophisticated defense concedes that readability alone is insufficient but argues that the source/binary distinction collapses for law: statutes, unlike compiled software, have “no binary layer” — the source is what executes, because law runs by being read and applied. This is the strongest analytical move available to the defense, and it fails on the crucial fact it omits.
Law is not compiled, true. But law is also not self-executing. An interpreted language still requires an interpreter, and the interpreter of law is not a machine — it is the human institutional apparatus: the judge who applies the statute, the prosecutor who decides to invoke it, the officer who enforces it, the jury that finds against a defendant. The claim that “the source is the executable” is true only in the trivial sense that the text is what the interpreter reads; it is false in the sense that matters, because the interpreter itself is the closed and unforkable part of the system.
The defense’s error is to conclude, from the absence of a compiled artifact, that source-available collapses into open. It does not. Reading the source of an interpreted language — Python, Perl, JavaScript — is not openness; it is source availability, and the interpreter still belongs to the incumbent. In open source, the distinguishing capability is that a user can obtain and run a copy they control: compile their own CPython, fork the reference implementation, observe behavior directly. The governed subject of Texas has no such capability. They cannot compile the Penal Code and run it; they cannot obtain a parallel court and ask it to adjudicate their version; they cannot fork a judge. The sole interpreter is the state’s, and its discretionary operations — charging, plea, jury, enforcement — are sealed against the governed.
The asymmetry is decisive. In open source, closed execution state does not foreclose openness, because the user can always obtain an instance they own. The governed subject owns zero instances; they are only ever operated upon by the state’s single instance, after the fact, through decision layers that are largely undisclosed. “The interpreter belongs to the incumbent” is not, as the defense would have it, a tolerable parallel to “CPython belongs to a core team” — because CPython is forkable by anyone with a compiler, while the state’s courts are not forkable by anyone at all. You cannot fork a judge.
4. The Maintainer Fallacy: The Users Are Not the Maintainers
The emotional core of the thesis is the identification of the governed with the maintainers: “citizens propose a law — that’s a pull request; the collective debate and votes — that’s code review and merge; the governed are the maintainers.” This is the inversion that the whole argument depends on, and it is false in both directions.
The governed are not the maintainers. The overwhelming majority of Texans will never author a statute, never draft a constitutional amendment, never merge a change to the state’s supreme legal file. The actual maintainers of Texas law are the incumbent organs: the legislature (which authors statutes and proposes amendments), the courts (which author common law and overrule precedent), and the executive agencies (which author administrative rules). These are components of the kernel, not the governed. The set of people who can author and merge the state’s source is a small, gated, incumbent-controlled class. The relationship between the governed and the law’s source is that of consumer to binary — a binary that a privileged few get to rebuild — not maintainer to repository.
The maintainer’s standard reply is that open source itself runs on tiny gatekept merge elites: Linux on a handful of subsystem maintainers with veto, CPython on a core team, rustc on a similarly narrow authority. This is true, and it is the defense’s great decoy. The reply does not answer the objection; it reframes it, and the reframing fails. The disanalogy is not that a gatekept elite exists in both systems — it does, universally — but that the gate is bypassable in one and not the other:
- In open source, the gate is a formality. Reject a contributor’s pull request and the contributor forks the repository and ships a divergent copy, unilaterally, needing no one’s permission. The maintenance elite is real, but it is conceptually irrelevant to the capability that defines openness, because forking does not require its consent.
- In Texas, the gate is a wall. There is no unilateral divergence of the state’s binding source. Every route by which law actually changes passes through the incumbent’s organs. The governed can neither author a constitutional amendment nor fork a statute to which they object.
The discriminator is therefore not the existence of the gate — which the defense correctly observes is universal — but its bypassability, which differs categorically. The defense has spent its energy proving that a gate exists in open source; it has never once answered the question of whether a bypass exists in Texas. It does not, and its absence is structural, not incidental.
5. The Amendment Fallacy: The Checkable Fact That Ends the Argument
The defense’s boldest claim is that Texas offers self-nomination to master: constitutional amendments go to a direct popular vote, requiring neither legislature nor governor — “a petition to the electorate merges the top-level file.” This claim is factually false, and its falsity is checkable against the state’s constitution.
The State of Texas has no statewide initiative or referendum, for constitutional amendment or for statute. It is one of a minority of states lacking any initiative process. The only route to constitutional amendment is a joint resolution passed by two-thirds of each house of the legislature, which is then submitted to the voters; there is no citizen-petition path, no signature drive, no way for a private party to place an amendment on the ballot without the legislature’s approval. A constitutional amendment that the incumbent legislature declines to propose can never reach the people; the governed never see it.
The consequence is that the incumbent has an absolute, un-waivable veto over what may be ratified. The people’s power over the state’s supreme legal file is a ratification veto — the power to approve or reject candidates the incumbent chose to float. This is the “pending approval” role in a protected branch: it can block, but it cannot compose, author, or push. In the defense’s own terms, the electorate holds a review ticket, not a push token. Self-nomination to master does not exist, because the sole authoring route is gated behind the incumbent’s two-thirds.
Two further corrections compound the falsity:
- The “master-branch governor” is a false fact. Texas constitutional amendments require no governor action whatsoever — no signature, no veto. The mechanism is two-thirds of each house plus voter ratification. The round-metaphor “the governor signs — master branch update” describes an operation that does not exist in the document it pretends to describe.
- The ratification vote is not a merge. Voting to adopt or reject an incumbent-floated amendment is the selection of a given diff, not the authoring of one. It is the difference between approving a proposed commit and composing a commit. The governed select from a menu the incumbent defines; they do not write the menu.
This single fact — the absence of statewide initiative — is the load-bearing wall of the entire argument, and it is where the thesis dies. Every route by which the governed might diverge the state’s source without the incumbent’s permission is closed; the one route that survives (ratification) is precisely the one that requires the incumbent’s permission to exist at all.
6. The Empirical-Fork Fallacy: Authorized Derivatives Are Not Forks
The defense’s most empirically loaded claim is that Texas is de facto forked: 254 counties, more than 1,200 cities, some 1,000 school districts, each “running local-variant bodies of code”; home-rule cities adopting charters that “deliberately diverge from the state default”; common-law courts “overruling” precedent to create divergent branches. The reply is that these are not forks — they are authorized derivatives within a permissioned franchise, and the distinction is categorical.
A fork, in open source, is a divergent copy that a party can produce and run regardless of the origin’s consent, and that can, in principle, compete with and replace the origin. The local governments of Texas satisfy neither condition:
- They are creations of the trunk. Texas created its counties, its cities, and its school districts; the state grants their existence and defines their powers. A home-rule charter that conflicts with the state constitution or the general laws of the state is void. Local “divergence” runs only to the extent the state’s own enabling statute permits, and it is revocably overridable: the state can preempt, dissolve, or absorb any local variant at will.
- They cannot fork upward. In open source, forking is bidirectional: a fork can out-compete the origin and replace the trunk, as Linux replaced Minix and as distributions have replaced one another. No Texas county can replace the state; no city charter can supersede the constitution; no local variant can absorb the trunk. Divergence in Texas runs only downward, and by permission. This is the plugin franchise — the trunk publishing an API that downstream variants may use within a sandbox — not a forking ecosystem.
The common-law instance is no better for the defense. Overruling precedent is divergence, but it is divergence exercised by the state’s own courts — kernel organs editing master, not users forking. Precedent is changed by judges, not by the governed; the parties who litigate submit inputs and await the kernel’s ruling, and they cannot themselves overrule. Every engine of legal change the defense cites — legislation, adjudication, overruling, administrative rulemaking — is staffed by the incumbent’s organs and exercised at their altitude. None is exercised by the governed, upward, against the source that binds them. The maintainer gap, established in section 4, thus re-derives at the constitutional, statutory, and administrative levels alike: at every altitude, divergence is performed by the kernel’s organs for the kernel’s subjects.
7. The Transactional Fallacy: A Contract Is an Input, Not a Fork
The defense’s most sophisticated relocation is to the transactional layer: the claim that private ordering — contracts, arbitration, choice-of-law — constitutes a permissionless fork marketplace, in which the governed bind themselves to customized rule-modules that the state’s courts execute openly. “Two parties sign a contract selecting foreign law; did they ask the state’s permission? No. Does the state execute it? Yes, by its own courts. That is divergence of the binding executable with no permission at any altitude.”
This is the most plausible defense on offer, and it fails on a single mechanical distinction: a contract is not a program the governed run; it is an input submitted to a program they do not control.
A contract is authored text that is filed and then processed by the state’s enforcement machinery. When a contract is enforced, the thing doing the executing is the state’s court, not the contract; the court applies its own rules of validity, its own procedure, its own sanctions, and it does so under the sole kernel’s authority. The parties do not compile and run their agreement — they submit it to an interpreter and hope the kernel schedules it. This is not running a divergent copy; it is requesting service from the only copy that exists.
The defense’s “permissionless” characterization confuses submission with execution. Submission is permissionless: anyone may write an agreement. Execution is not: the kernel decides whether the module loads, and it decides on the basis of its own validity doctrines — public-policy voidness, jurisdiction limits, illegality, unconscionability. Parties cannot contract into slavery, cannot contract out of criminal liability, cannot strip the courts’ jurisdiction, cannot enforce an illegal arbitration. The “fork” is therefore valid only within the sandbox the kernel’s doctrine already drew, and the moment the kernel narrows or declines, the agreement evaporates — it was never running; it was pending the interpreter’s pleasure. What the defense calls a permissionless fork is in fact revocation-contingent permission: the fork runs only while the kernel declines to notice, and the kernel always retains the power to notice. A system that can void your divergence is a system that permitted your divergence — and permission, being granted, is not freedom.
The choice-of-law instance confirms, rather than rebuts, this structure. The state does not enforce foreign law; it enforces its own conflict-of-laws rule, which directs its own courts to apply a specified body of rules in a specified situation. The parties choose from a menu the kernel itself defined; the enforcement act remains the kernel’s. This is not forking the executable; it is selecting an argument to an API whose dispatch table the kernel controls. The defense’s error is to mistake the kernel’s own menu for a window the parties opened.
Finally, the transactional defense re-imports the vacuity it spent its earlier rounds trying to escape. Private contracting exists in every jurisdiction with a market — the People’s Republic of China enforces contracts daily, Riyadh’s commercial courts apply choice-of-law clauses, joint ventures in Pyongyang select foreign law and arbitral seats. If “the governed can bind themselves to customized rule-modules enforced by courts” is the test, then every market-state passes it, and the test discriminates nothing. The defense that anchors openness in the transactional layer has relocated the discriminating work to a layer that is forkable everywhere — and thereby proved the opposite of what it set out to prove.
8. The Judicial-Appointment Fallacy: Personnel Are Not Architecture
The defense’s most novel move is the claim that the governed can obtain the interpreter: “You cannot fork a judge — but you can become one.” Texas district and appellate judges are elected, district attorneys and sheriffs are elected, and jurors are drawn by lot from the governed. Therefore, the defense argues, the interpreter is obtainable by the governed in a way that the hardware beneath all software is not.
This collapses a fundamental distinction: personnel are not architecture. Becoming a judge does not obtain an instance of the interpreter; it occupies one seat in the sole kernel that already exists. There is still exactly one court system, and it is unforkable. A new judge does not own a divergent copy of the law; they become a component of the only copy there is — bound by the very statutes they cannot alter, constrained by precedent, subject to appellate review and, ultimately, to recall. A judge’s discretionary output is the kernel’s output, not a fork of the kernel. Being part of a system is categorically distinct from owning a fork of it, and the defense’s move is precisely this category error: it treats enlistment in the incumbent apparatus as access to a divergent instance.
The disanalogy with open source survives the membership gate. Forking CPython requires no one’s consent: it needs a repository, a compiler, and ability — and it is open to any person on Earth. Obtaining a judicial seat requires winning an election — a process run by existing political machinery (money, party organization, geography, turnout) that is emphatically not lawful self-nomination, and that yields, at the end, a role inside the one unforkable copy rather than ownership of a divergent one. The governed cannot obtain the interpreter in any sense that parallels forking; they can, at best, apply to operate it, subject to incumbents’ machinery, and their tenure leaves the source as closed as it found it.
9. The Coercion Fallacy: The Structural Line Between Opt-In and Subjection
The defense’s least technical but most emotionally weighted claim is that coercion cannot disqualify Texas, because all open source rests on state coercion: every open-source license is enforceable only through copyright law, courts, and injunctions, so “must be non-coercive” would rule out every open-source project, including the ones the Texas thesis is modeled on. The move is fair as far as it goes, and it must be conceded at its root: coercion underlies all software licensing, and “is coercion present?” cannot be the discriminator, or the term would be vacuous for software as well.
But the discriminator was never the presence of coercion; it is the consent structure of the coercion — and there the two systems differ categorically, not by degree.
Open-source coercion enforces an opt-in, scoped, voluntary contract with a real outside. You chose to download and use the software; the license terms attach to an exchange you engaged; and — the decisive feature — you can decline. Decline to redistribute and the coercion ends: you have made a complete, exercised exit. There is an actual outside to the GPL’s obligations, and you can occupy it.
State coercion is structurally the opposite. It applies involuntarily to everyone within a territory, on the sole basis of existence in that place, whether or not they engaged any exchange, signed any agreement, or consented to any term. It is unlimited in scope — it attaches to everything within the territorial jurisdiction, and its reach is retroactive and prospective without any act by the subject. And it has no outside within the inhabited world: leaving Texas is only ever entering another state that coerces you on the same non-optional terms. Escape from the GPL is real and complete (decline redistribution). Escape from a state is only ever lateral — into another instance of the very non-optional coercion you cannot decline, because the set of jurisdictions has no complement within the inhabited world that you can occupy.
The defense is right that “relocation exists.” The reply is that relocation is not exit; it is a lateral move within a closed set, and the set is closed because its members are non-owned and non-forkable. In open source, the lateral move (Windows to Linux to BSD) is a move among user-forkable, user-owned copies — the code you then run is yours to fork at will. In law, the lateral move (Texas to New Mexico) is a move among prefabricated, non-owned kernels — you switch which closed instance you are subject to, but you never fork the instance itself, and you can only ever select from the ~200 that already exist. A complement-free set of forkable members is an open universe; a complement-free set of non-forkable members is a closed one. Identical topology, opposite openness. The coercion defense does not dissolve the fork test; it accidentally supplies the contrast that confirms why the law fails it.
10. The Definitional Fallacy: Every Honest Standard Excludes Texas
The defense’s last redoubt is to redefine the term into compliance. Every redefinition attempted takes one of two forms, and both fail.
Form one: a test narrow enough to exclude the authoritarian states but broad enough to admit Texas. The canonical candidate is a conjunction: a system is open when its executable source is public AND amendable without incumbent veto AND its interpreter is obtainable by the governed AND its primary-function extension interface is open AND its refusal layer is authored rather than immutable. This has the correct shape of a definition — it is conjunctive, and it discriminates. But a conjunction is only as strong as its weakest predicate, and Texas fails the one predicate that does the discriminating work: amendable without incumbent veto. As established in section 5, Texas is not amendable without the incumbent’s veto — the two-thirds legislative gate is an absolute, un-waivable gate over what may ever be ratified. The defense’s own definition, honestly applied, excludes Texas. And if the defense retreats this conjunct to mean merely “a ratification vote exists,” it has reopened the vacuity it built the conjunction to close: Beijing could add a referendum to its constitution and pass the whole test tomorrow.
Form two: a test broad enough to admit Texas, which thereby admits everything. Any definition of “open” in terms of published text, or modifiable-in-principle text, or contract-level divergence, is satisfied by every market-state and every authoritarian jurisdiction with a statute book. A test that admits Beijing, Pyongyang, and Riyadh is not a definition of open source; it is a definition of “has written law,” and it concedes the question by refusing to discriminate.
The two forms exhaust the space. A test narrow enough to exclude the states the proponent does not wish to call open source excludes Texas on its own terms; a test broad enough to admit Texas admits them all. There is no third form, and the defense’s repeated oscillation between the two — narrow in one round to claim exclusivity, broad in the next to claim inclusion — is the signature of a definitional position that cannot be consistently maintained.
11. The Historical Fallacy: Founding Forks Are Not Forking
The defense’s most romantic claim is historical: Texas was “forked from Mexico’s legal runtime” in 1836, forked again in 1861, and its 1876 constitution has been amended roughly five hundred times since — “a repo forked continuously since 1876, and the repo never went private.”
The historical record refutes the romanticism on its own terms.
- The founding “forks” were imposed, not achieved. The 1836 republic was not created by the governed forking a permissive source; it was asserted through armed conflict and consolidated by a convention of political elites. The 1876 constitution was drafted by a convention, not ratified by an open pull-request pipeline of the governed. A fork that can only be achieved by overthrowing the previous maintainers, and that is then operated by a self-selected convention, is the failure of a fork mechanism, not its provision. Open-source forking is peaceful, cheap, and lawful — the very opposite of the armed restitution that founded the state.
- The 1861 fork was forcibly merged back. The state’s “fork” into the Confederacy was re-merged at gunpoint, without the consent of the governed, in the course of a war. This is the license failing, not the license working. It is the open-source equivalent of a maintainer rewriting history and forcing a re-merge — an operation that no open ecosystem tolerates, because it is the elimination of divergence, not its facilitation.
- The roughly five hundred amendments are all incumbent-referred. As established in section 5, not one of them reached the ballot except through a two-thirds vote of both houses of the incumbent legislature. A “commit history” in which every commit must first be drafted and approved by the maintainers, and in which the governed may only ratify or reject maintainer-floated candidates, is not an open contribution pipeline. It is a closed authoring model with a public ratification step bolted on.
The historical argument, examined, is not evidence of forking. It is evidence of the precise structure the rest of this essay establishes: a closed authoring kernel, occasionally ratified by the governed, and — on the two occasions the governed did seize the source — seized by force, not by fork.
12. The Metaphorical Claims: The Jury, the Court, the Vote
The roundest defenses extend the software metaphor to specific institutions: the courts striking down a law are “a revert of a bad commit”; the jury is “the test suite that checks the system on real inputs”; voting is “the release/rollout process — you approve the build.” Each is a selected feature of the metaphor, and the selection is telling: every chosen feature is one in which the governed have a passive, reactive, or negative role, and every omitted feature is one in which the governed would need an active, authoring, or forking one.
- The court as revert. A court striking down a statute does not revert a commit; it refuses to execute one, and the refusal is performed by a kernel organ, not by the governed. The statute remains on the books until the legislature — a kernel organ — acts. Moreover, the court’s power to refuse is itself a kernel feature the governed cannot invoke on demand; they can only petition the kernel to exercise it.
- The jury as test suite. The jury does not run the system; it is run by the system. Its deliberations — the part that would make it an interpretive engine — are sealed. The jury is a veto held by the governed over the kernel’s application to a specific case, exercised under rules the kernel defines, and it decides only inputs the kernel decides to submit. A “test suite” that can only reject single inputs the maintainer chooses to run, and whose reasoning is sealed, is not a test suite; it is a quality gate the kernel owns.
- The vote as release process. Voting selects operators of the compiler; it does not author the source. The governed can replace a politician — re-roll a component of the kernel — but they cannot change the rules the kernel executes, and they cannot fork them. Selection of personnel is not control of source. The vote is the governed’s power to choose who compiles; it is not a power to choose what is compiled or whether they may compile their own version.
The metaphor fails because it is an aesthetic, not a mechanical account. It collects the features of the state that resemble some aspect of a software project and omits the one feature that defines open source: the governed’s unilateral power to diverge and run the source. A metaphor that must discard the term’s defining property in order to sustain the comparison is not an argument; it is a selection effect.
13. The License-Axis-Only Defense: The Narrow Claim, Ruled
The strongest remaining defense is also the narrowest: abandon the fork axis entirely and claim that on the literal OSI license axis alone — free redistribution, source in preferred form, permitted derived works, nondiscrimination, technology neutrality — the public-domain statute book clears the definition. This claim is granted, for what it is, and it proves nothing.
Public domain statutory text satisfies most of the literal license criteria: it is freely redistributable and reproducible (confirmed as a matter of federal constitutional doctrine by Georgia v. Public.Resource.Org, holding that official law is uncopyrightable because every citizen is presumed to know it); it is “source in preferred form” (the statute is the source); derived works are permitted maximally (public domain is the most permissive regime that exists); commercial use is unconstrained. The narrow, literal, license-only claim is true.
But it is true for the reason this essay has maintained from the outset: the license axis was never in dispute. The verdict “source-available” has always designated precisely this — a maximally permissive license axis, freely redistributable text, modifiable in principle — wrapped around a fork axis closed to the governed. The license-axis-only defense does not rescue the thesis; it is the “available” in “source-available,” the half that was conceded before the argument began. And it is undone by the coupling established in section 1: the OSD’s license criteria exist to guarantee fork, build, and run; a license whose permissions the architecture cannot discharge is a promise the machine refuses. To claim “open” on the license axis while the fork axis is closed by an incumbent gate is to claim that a document is a contract when the only party who can perform under it has declined to perform, and the only other party in the room is the one who drafted it.
The narrow claim is therefore ruled as follows: granted as a statement about the license status of statutory text; rejected as a submission that Texas is open source. The two are different claims, and only the first survives — and the first is the definition of source-available, not open source.
14. The Verdict: Source-Available, Fully Specified
The argument across every possible axis converges on a single, complete, mechanical account of the verdict.
The license axis discriminates nothing. Every statute book on Earth — Texas’s included — is freely redistributable and modifiable in principle, because law is uncopyrightable public domain. This is the “available” in “source-available,” and it was never in dispute.
The fork axis discriminates everything. The difference between open source and source-available is not whether the text is public, not whether a maintainer elite exists, not whether coercion is present — it is whether the gate can be bypassed. In open source, the gate is a formality: rejected contribution or not, a third party forks and ships, unilaterally. In Texas, the gate is a wall: there is no path to diverge the state’s binding source without the incumbent’s approbation, because the only amendment route runs through a two-thirds veto of the incumbent legislature, the only un-permissioned divergence (a contract) is executed at the kernel’s pleasure and subject to the kernel’s voidance, and the only “interpreter” access available to the governed is enlistment as a component of the sole unforkable kernel.
The complete mechanical account is this:
- The top-level legal file is sealed behind an incumbent authoring gate (the two-thirds legislative veto; no statewide initiative). The governed hold only a ratification veto over incumbent-floated candidates.
- Every engine of legal change — legislation, adjudication, overruling, administrative rulemaking — is staffed by kernel organs, exercised at their altitude, not by the governed, upward, against the source that binds them.
- The sole interpreter of the law — the human coercive apparatus — is unforkable: the governed cannot obtain a parallel instance, and its discretionary execution (charging, plea, jury, enforcement) is sealed against them.
- The only un-permissioned divergence available to the governed is a contract, which is not a fork but an input — revocation-contingent, executed at the kernel’s pleasure, valid only within the validity doctrines the kernel authored.
- Coercion applies involuntarily and territorially to subjects with no outside within the inhabited world — a closed set of non-owned, non-forkable members, which is the opposite of an open universe.
Each of these facts is independently falsifying; together they are exhaustive. There is no route the governed can take to diverge the state’s source and run a modified copy without the incumbent’s permission, and every route they are offered — ratification, election, contract, petition — is a veto, a selection, or an input, never a fork.
Conclusion: The Gate Is the Wall, Not the Formality
The claim that Texas is open source is false, and it is false with a fully specified mechanism rather than a bare denial. The state’s legal text is public and freely redistributable — the most permissive license axis that exists. Its enforceable source is closed to the governed — a fork axis sealed behind an incumbent gate no subject can bypass. The most open license axis in existence is wrapped around the least open fork axis that could be built, and the two halves do not conflict; they are the reconciled verdict, and they agree.
Texas is source-available: a maximally permissive license on a system whose execution belongs to an incumbent that its subjects cannot fork, cannot bypass, and cannot exit into any place that is not itself such an incumbent. That is not open source with an unfortunate build constraint; it is the exclusionary definition of open source’s negating case. The gate is the wall, not the formality — and on that single, checkable, structural fact, the thesis ends.
The answer to the question is no. The mechanism of the no is now complete.