Skip to content

Infrastructure status policy

CategoryInfrastructureStatusDraftLast reviewed2026-09-12

This policy controls how MoreBC2 describes public BitcoinII / BC2 infrastructure.

It is meant to prevent dated observations, project-linked hostnames, or familiar API shapes from turning into broader claims about uptime, official operation, backend independence, wallet support, broadcast support, or custody suitability.

Use the narrowest accurate label.

Label Meaning
Observed A page, endpoint, or service responded during a dated check.
Locally Tested A local BitcoinII Core command or behavior was exercised in a documented environment.
Same-time comparison Two or more sources were compared during a recorded time window.
Partial Useful evidence exists, but major behavior remains untested.
Unavailable The service or endpoint failed in the recorded check.
Unknown No current committed evidence is available.
Historical Previously relevant evidence that should not be treated as current service state without a fresh check.
Roadmap Planned or expected, but not current evidence.

Service role and service health are separate claims.

Use role labels only when the evidence supports them:

  • Official — a project-controlled or service-controlled source explicitly designates the service as official.
  • Project-linked — the hostname or project documentation links the service, but that does not prove project operation.
  • Independently operated — the service itself or another sufficiently direct source states independent operation.
  • Supplemental — useful third-party/public infrastructure without official-project designation.

A hostname under bitcoin-ii.org is not, by itself, proof that the BitcoinII project operates the backend.

Do not imply that a service is:

  • permanently synced,
  • reliable,
  • official,
  • project-operated,
  • independently redundant,
  • safe for custody,
  • suitable as an exchange’s only source of truth,
  • wallet-compatible,
  • valid-transaction broadcast-compatible,
  • rate-limit stable,
  • operated by a specific maintainer,
  • covered by release or binary verification,

unless a committed evidence record directly supports that exact claim.

Live services can change quickly. Every service claim should include, where applicable:

  • last checked date,
  • evidence source,
  • what was checked,
  • what was not checked,
  • known limitations.

If those fields are missing, use Unknown, Historical, or an equivalent qualified label rather than presenting the service as currently active.

A successful dated probe is evidence for that probe window only.

Use:

matched at check time

or:

reported the same height and best-block hash during the recorded comparison window

Do not automatically use:

synced

A same-time match is valuable evidence but does not establish permanent synchronization, shared chain-selection behavior under every condition, or long-term reliability.

Multiple hostnames do not automatically equal multiple independent providers.

Before describing services as independent redundancy, look for separate evidence about:

  • operators,
  • node backends,
  • databases/indexers,
  • hosting/network providers,
  • failure domains,
  • service-control paths.

Closely aligned route shapes, schemas, values, WebSocket behavior, or error behavior may justify a caution that independence is not established. Those similarities do not, by themselves, prove a shared backend.

Transport success and application success are separate.

Examples:

  • HTTP 200 with success:false means the route is reachable but the requested function did not succeed.
  • HTTP 404 is evidence that the tested route was unavailable at that path/time.
  • HTTP 500 is evidence of server-side failure for the tested request, not necessarily permanent service failure.
  • A parameter-validation error can establish route presence without establishing successful operation for a complete request.

Document both the HTTP result and the application-level result when they differ.

Use strict wording for broadcast routes.

If an endpoint accepts POST and rejects deliberately invalid transaction data, safe wording is:

transaction-submission route exists and rejects malformed input

That does not support:

valid BC2 transaction broadcast tested

Successful valid-transaction broadcast requires an actual valid BC2 transaction test under an explicitly safe test plan and must be recorded separately.

Successful server.version, TLS, or header calls establish protocol reachability for those methods only.

They do not establish:

  • wallet compatibility,
  • address-history correctness,
  • fee behavior,
  • transaction construction,
  • transaction broadcast,
  • long-running subscription stability,
  • reconnect/load behavior.

Public explorer/API services are useful for observation and cross-checking, but MoreBC2 should not present them as replacements for an operator’s own BitcoinII Core node in critical deposit, withdrawal, custody, or chain-risk workflows.

A service provider may use public infrastructure as secondary monitoring while retaining its own node as the primary authority.

Maintainer/operator statements can be recorded as context, but they should not be converted into unrelated technical evidence.

Example:

  • acceptable: “The service identifies itself as independently run/community-funded.”
  • not acceptable: “The service is independently redundant at the backend level.”

Similarly:

  • acceptable: “The maintainer has indicated release signing is roadmap work.”
  • not acceptable: “Release signing exists.”

Status: Draft
Primary sources checked: September 2026 infrastructure evidence, current MoreBC2 evidence language, and linked verification records
Notes: This policy is editorial guidance. It does not independently verify any external service.