What Open Source Means for ViaSign

Open source can sound simple: make a repository public so other people can read, run, question, or improve the code.

For ViaSign, that is not enough. The project works near language, accessibility, and community knowledge. Making code visible may support accountability, but it does not make the code linguistically correct, accessible, or trusted. It does not replace the people whose language and lived experience the system is trying to respect.

The boundary I want open source to hold is this:

Open source can begin accountability, but it does not create linguistic or community authority.

Why I published a smaller boundary

Some ViaSign research began privately while I was still trying to understand the problem. It explored questions around English input, SgSL structure, review, sign planning, and motion.

Publishing that whole workspace would not automatically make it responsible. Private research can contain uncertain assumptions, mixed provenance, and experiments that look more complete than they are. Some examples were never intended for redistribution.

So the public ViaSign parser was rebuilt as a smaller clean-room boundary. It is not a copy of the private parser history, and private benchmark material was not migrated into it. Its purpose is to make a limited contract visible without pretending the wider system is ready.

That is less dramatic than publishing a large demo. It is also more honest.

What the public parser does today

The current public parser performs limited natural-English surface analysis. It can return a few closed categories while keeping uncertainty explicit. It does not implement SgSL grammar, translate English into SgSL, produce sign plans, or generate motion.

Two invariants remain visible in every ordinary result:

review_required = true
motion_ready = false

Those values are part of the public contract. They stop a small amount of surface analysis from being presented as approved signing or motion-ready language.

This is why I describe the parser as a safety boundary, not the completed ViaSign grammar system. The public ViaSign boundary separates what exists now from research that may come later.

Inspectability is not authority

Public source gives people something concrete to inspect. A developer can read the contracts, tests, and failure states. Someone can ask why a category exists, whether uncertainty is visible, or whether a response crosses the documented boundary.

That matters, but engineering inspectability is not linguistic validation. A passing test suite cannot prove that a future grammatical decision reflects how signers use SgSL. Public code cannot claim Deaf community endorsement simply because Deaf people are able to see it.

I bring my own Deaf and hard-of-hearing lived experience to ViaSign, but I do not represent every Deaf person, signer, teacher, learner, interpreter, or SgSL user. ViaSign still needs relationships and review processes that do not exist formally yet. The community review process describes a direction to build toward, not a claim of completed validation.

Open source can make correction possible. It cannot decide whose correction carries linguistic or community authority.

Responsible openness includes boundaries

The parser’s published code and original documentation are available under the Apache License 2.0. That licence does not grant rights in Singapore Sign Language itself. It does not make private lesson material, community knowledge, personal data, recordings, or project trademarks available for reuse.

Language and community knowledge are not raw material waiting to be placed in a repository. They can carry lived experience, teaching, correction, and cultural knowledge that should not be treated as free data to extract.

Responsible openness cannot mean putting everything online. Sometimes the accountable decision is to show the code. Sometimes it is to explain why a boundary exists. Sometimes it is to keep a source private until permission and provenance are clear.

Contribution should remain reviewable

The public repository includes contribution and conduct guidance. It asks for small, test-backed, public-safe changes that preserve visible uncertainty. It also asks people not to submit private lesson material, community media, personal review data, or claims of final SgSL correctness.

This is a starting point, not a mature contributor programme. There is no public release or formal support promise yet, and I should not describe future participation as though established review relationships already exist.

If people contribute later, their work should be attributed and respected. A technical contribution should be reviewed as engineering work. A linguistic claim should require relevant knowledge and review. Lived experience should not become an unpaid approval step for a decision already made.

Public source is not a public service

The repository is public. Public parser access is not.

The separate FastAPI service was deployed to Render and was last verified in maintenance mode with empty CORS. Its live provider state must be reverified before activation. The production Viathorne website is not configured to call it, and the tester interface remains disabled.

Publishing source, deploying inert infrastructure, and accepting public input are different decisions. A repository can be safe to inspect before a service is safe to operate. Technical checks also cannot replace the linguistic, accessibility, privacy, and community review needed for activation.

The Build Log records these checkpoints because “the code is public” should never silently become “the system is ready.”

What still has to come later

SgSL grammar, sign planning, motion, avatar work, and formal community review remain future work. Each layer needs its own evidence and review gate.

Future grammar rules should show their sources, assumptions, uncertainty, and correction path. Future motion should remain traceable to reviewed decisions rather than hiding them behind smooth animation. Community participation should be described with the consent of the people involved, not used as a broad endorsement claim.

Open source may help those layers become inspectable. It cannot replace relationships, teaching, language knowledge, or the willingness to change direction when an assumption is wrong.

That is why review_required should remain more than a prototype label. It is a reminder that uncertainty belongs inside the system, not outside it as a small disclaimer after the output has already appeared confident.

Accountability before attention

I do not want to open source ViaSign just to make the project look active or complete. I want the public parts to be small and clear enough that their limits can be questioned and their mistakes can be corrected.

That means showing what the parser does. It also means showing what it does not do. It means separating public code from private sources, engineering review from linguistic authority, and future participation from relationships that have actually been built.

Trust is not a label an open repository can give itself. It has to be earned through behaviour: clear boundaries, evidence before claims, consent, visible uncertainty, and correction when the work is wrong.

For ViaSign, open source is one way to begin that work in public. It is not the end of the responsibility.