Nanopublication versions and retractions
How a nanopublication is replaced or withdrawn, why only the same key counts, and how deriving and overriding let others build on or revise it.
Why there are versions
A nanopublication cannot be changed after it is published: its address is computed from its content, and copies of it sit on several servers. So correcting a mistake, updating a definition, or withdrawing a claim is done by publishing another nanopublication that refers to the first one. The original stays on the network, but every tool learns to treat it as outdated.
There are two such moves: publishing a new version, which replaces the old one, and retracting, which withdraws it without replacement.
New versions
A new version is a complete nanopublication of its own that names the one it replaces in its publication info with npx:supersedes. Tools then show the new one wherever the old one used to appear, and mark the old one as superseded.
Versions form a chain: the newest one supersedes the previous, which superseded the one before. Any address in the chain can be resolved to the current version, so a link to an old version keeps working and leads to the latest state. Nanodash offers "supersede" on any nanopublication, which opens the form pre-filled with its content.
Retractions
A retraction is a small nanopublication whose assertion says, with npx:retracts, that a given nanopublication is withdrawn. The retracted one remains on the network and can still be opened, but it disappears from listings and search results, and anything it introduced no longer counts.
Retracting is the right move for something published by mistake or found to be wrong with nothing to put in its place. If there is something to put in its place, publish a new version instead; that both withdraws the old one and provides the replacement.
Only the same key counts
In principle anyone could publish a nanopublication claiming to supersede or retract another, so tools apply one rule: a new version or a retraction is honoured only if it is signed with the same key as the nanopublication it refers to. Otherwise it is stored like any other nanopublication but has no effect on the original. The command-line client will not even build a retraction under a different key.
Two consequences follow. A nanopublication signed with a key that is lost can no longer be superseded or retracted by that route. And you cannot revise someone else's nanopublication this way, however wrong it is — for that, see deriving and overriding below.
What is introduced keeps its identity
Many nanopublications introduce something — a term, a role, a view — and give it a URI minted under the nanopublication's own address, recorded with npx:introduces. When such a nanopublication is superseded, the new version must keep that URI exactly, rather than mint a new one under its own address, so that everything referring to the thing keeps pointing at it.
The form does this automatically when opened in supersede mode. When writing a new version by hand, this is the one thing to get right: copy the introduced URI from the original, do not let it be regenerated.
Deriving: a new thing based on an old one
Sometimes the aim is not to replace a nanopublication but to make a different one that builds on it: a variant of a definition, an adaptation of a template, a claim narrowed to another case. This is a derivation. The new nanopublication records prov:wasDerivedFrom pointing to the old one, mints fresh identifiers for what it introduces, and leaves the old nanopublication as it is.
Since nothing is replaced, deriving works across keys: anyone may derive from anyone's nanopublication. Nanodash offers "derive" next to "supersede".
Overriding: revising someone else's work
An override sits between the two. Like a derivation it records prov:wasDerivedFrom and can be published with a different key; like a new version it keeps the identifiers of what the original introduced. The result is an in-place revision of the same thing by a different person.
Whether an override takes effect is not decided by the key but by convention on the page where the thing appears: pages maintained by a community typically show the newest version published by any member of that community. This is how shared content — documentation, vocabularies, views — is kept up to date by a group rather than by whoever happened to publish it first.
Forks
If two people, or one person twice, publish a new version of the same nanopublication independently, the chain forks: two nanopublications both supersede the same original, and tools cannot tell which is current. Pages then tend to keep showing the version they had, or pick one at random.
A fork is repaired by publishing one more version that supersedes both heads. To avoid causing one, check what the current version is before superseding, especially for nanopublications that several people maintain.
How the current version is found
Tools find the current version by following the chain of npx:supersedes links forward from any version, staying with nanopublications signed by the same key, and stopping at the last one that has not itself been superseded or retracted. Nanopub Query provides this as a service, so every tool resolves versions the same way.
For things governed by a space, an alternative rule applies alongside: the current version is the newest one published by a member of that space, whatever key signed it. That is what allows a community to carry a definition forward after its original author has left.
Good practice
Before superseding, make sure you are superseding the current version and not an older one. Keep the introduced identifiers. Add a short change note when a new version is more than a typo fix; the form offers a field for it. Retract promptly when something is wrong and there is no replacement, since the retracted nanopublication is otherwise taken at face value.
And do not fear the permanence: a superseded or retracted nanopublication is hidden from everyday use, but it remains available to anyone who cites it, which is exactly what citing something should mean.
This page is a copy of the topic Nanopublication versions and retractions of the Nanopublication Docs. Each paragraph is a nanopublication. In Nanodash you also find their examples, related paragraphs, and version history, and members of the nanopub space can edit them there.