Composing With OriginTrail: A Thesis With No Code Behind It, Said First Rather Than Last

There is no adapter, and that is the first fact rather than the last

We ship four framework adapters. This is not one of them.

Measured: their name, their protocol vocabulary and their token return nothing at all across both published packages and the backend. The identical search finds the four vendors that do have adapters across four files, so the searcher works.

So everything below describes a shape that would fit, not an integration that exists. A page that put that sentence at the bottom would be a different and worse page.

Why the shape genuinely fits, argued from what each thing answers

They record provenance. Where a piece of data came from, what produced it, what happened to it.

We issue credentials. Who an actor is, who is accountable for it, and what it is permitted to do.

Those are different questions and they are adjacent in the useful way. A provenance record that says "this output came from agent X" is more valuable when something checkable can be said about agent X, and a credential about an agent is more valuable when what the agent actually did is recorded somewhere durable.

That is the whole argument for composing, and notice that it needs no claim about their gaps.

What composing would concretely mean

A credential reference travelling inside a provenance record, so a consumer reading the record can verify a statement about the actor without an account with either of us.

What makes that possible on our side today: the credential format is a published standard and the identifier method is registered in the W3C DID Method Registry, so verification needs the issuer's key and none of our code.

And what blocks it today, all on our side

No resolver driver exists for our identifier method, so a consumer of a provenance record who tried to resolve the identifier through standard tooling would get nothing. A reference nobody can follow is a string.

And the registry slot where evidence of this kind belongs is one we do not fill, because our validation configuration points at nowhere pending an upstream deployment.

Three blockers, and none of them is theirs.

What our own note says that this page will not print

Named enterprise partners, a market figure, a membership count and a scoring table. We verified none of them, and a secondhand record about another company is a hypothesis rather than a source.

Their token, which belongs to a different conversation than this one entirely.

What the note supports, stated as what it is: a dated record that they publish on the standard this surface anchors to, and that their published material is the template we would model an integration on. That is a note from May 2026, not a partnership, and nobody has agreed anything.

What is already built, because absences alone would misdescribe us

A credential format that is a published standard, verifiable by any conformant implementation holding the issuer's key.

An identifier method registered in the W3C DID Method Registry.

A live passport in a public registry, readable without an account.

And four adapters that demonstrate the wrapper pattern an integration here would follow, so the work is a known shape rather than a research project.

And a discipline this page is deliberately keeping

Our shipped adapters carry characterisations of other companies' products in their source comments, and this estate filed that as a defect rather than treating it as normal.

So this page describes our side and quotes a dated note for theirs. If an adapter is ever written here, its comment should say what our code does and nothing about what theirs lacks, because a comment ships in a tarball and cannot be edited afterwards the way a page can.