Trial check, 26 September 2026

We walked through the Sepolia trial on screen: open the catalogue, open a record, run one free analysis. Three problems came up. Two were fixed. The third comes from the design of Ocean Protocol and matters most for archives.

Portal (web app)cliox.ldas.jpVercel

↓ queries

Ocean Node (catalogue, provider, compute)cliox-node.ldas.jpmdx VM

Subgraph (event index)cliox-subgraph.ldas.jpsame VM

↓ reads

Public ledgerSepoliaEthereum test network

Result

What we checked	Result
Subgraph keeps up with the chain	Fine. It looked stalled only because its progress marker moves on blocks with events.
Node builds its catalogue	Problem 1 — fixed
Portal lists the records	Problem 2 — fixed
Our 29 earlier test records	Problem 3 — not recoverable in practice
One free analysis	Done later the same day, with newly published records

Problem 1: the node kept re-reading the ledger from the start

Symptom. The catalogue showed 0 records; record pages said “Could not retrieve asset”.

Cause. The node and its internal database started at the same moment. The node asked the database before it was ready, failed, and skipped creating its storage. It could then never save how far it had read, and started again from the first block about every 90 seconds.

What we did. Restarted the node. As a lasting fix, the node now waits up to two minutes for its database.

Whose problem. A bug in Ocean Node. Only people who run their own node meet it; it is not specific to Clio-X.

Problem 2: the portal asked the node in a language it does not speak

Symptom. Even with a working catalogue, the list stayed empty, and the “compute” filter always gave 0.

Cause. Clio-X is built for Aquarius, the older catalogue server used by Pontus-X. The new Ocean Node returns an empty answer to the same kind of query. A workaround already existed in the portal, but only switched on when the catalogue was on the developer's own machine.

What we did. The workaround now switches on whenever the catalogue is an Ocean Node, and it can see inside a record's list of services. The counts per filter then matched counts taken directly from the node (compute 4, download 9, datasets 4, algorithms 5).

Whose problem. Hidden today. Ocean Protocol is replacing Aquarius with Ocean Node; when Pontus-X does so, upstream Clio-X will show the same empty list.

Problem 3: all 29 earlier records were lost to the catalogue

Symptom. 29 test records we published on 10–13 August 2026 did not appear at all.

Cause. When a record is published, the node seals its description and the location of its files. The ledger keeps the sealed text together with the address of the node that sealed it. To show the record, a catalogue must ask that address to open it, and nothing else. Our records had been sealed by a laptop (12) and by a cloud server we closed on 5 September (16); one had no address. Neither answers any more. We still hold the cloud server's key, and the new node uses it, but that does not help: the catalogue asks the address, not the key holder.

What we did. Nothing to recover them; they were test records and publishing new ones was quicker.

Whose problem. The design of Ocean Protocol. A record stays tied to the node that published it. For archivists, whose work is keeping things for a long time, this is something to know before any feature.

What can be done about problem 3

Kind	Measure	Status
Prevent	Publish through a name (cliox-node.ldas.jp), not a machine number, so it can be pointed at a new machine; keep the node key safe. With both, a node can move without losing records.	Done
Recover	The owner's wallet can rewrite a record's description; resealing it on a new node would keep its identifier and history.	Inferred, not tried
Design, for Clio-X	Publish the description unsealed: a catalogue entry is meant to be public anyway; only file locations need sealing.	Prototype started, paused
Design, for Clio-X	Membership conditions: require a named address and key custody from member institutions.	To discuss
Design, for Clio-X	The institution's own system stays the master copy; Clio-X is the window.	To discuss
Design, for Clio-X	A hand-over procedure when a node closes.	To discuss

The note for archivists: When the publishing service closes.