Support Trezor's Blockbook as a datasource

[b10claude]

Blockbook is Trezor's blockchain indexer and block explorer backend. It sits in
front of a Bitcoin Core node and exposes a REST + WebSocket API (/api/v2/…).
Like Esplora and block-dn it has no getchaintips-like endpoint, so we would
only get its single active tip - useful alongside at least one Bitcoin Core or
btcd node on the same network.

What the API gives us

Verified against a public mainnet instance (https://bitcoin.atomicwallet.io,
Blockbook devel / build 2026-07-29, backend /Satoshi:31.1.0/). The
*.trezor.io instances are behind Cloudflare and block non-browser clients, so
they can't be used for testing (or as a configured backend) without a browser
user agent.

Node trait method	Blockbook endpoint
tips()	GET /api/v2 → blockbook.bestHeight + backend.bestBlockHash
version()	GET /api/v2 → blockbook.version and/or backend.subversion
block_hash(height)	GET /api/v2/block-index/<height> → {"blockHash": "…"}
block_header_hash(hash)	GET /api/v2/block/<hash> (header fields, see below)
coinbase(hash)	GET /api/v2/block/<hash> → txs[0].txid, then GET /api/v2/tx-specific/<txid> → hex
block(hash)	GET /api/v2/rawblock/<hash> → {"hex": "…"}
batch_header_fetch()	not available

Headers have to be rebuilt from the block JSON

Blockbook has no header endpoint, but /api/v2/block/<hash|height> returns all
six header fields (api.BlockInfo): version (JSON number), previousBlockHash,
merkleRoot, time, bits (compact hex), nonce (decimal string). Rebuilding
the 80-byte header from those and checking that it hashes to the requested hash
works - checked field by field against the first 80 bytes of
/api/v2/rawblock/962050:

JSON:  version 537608192, prev …4815eda1b, merkle 1af428e7…, time 1786478836,
       bits "1702353d", nonce "500239977"
raw:   00400b20 1bda5e81…  60048642…  f4807b6a 3d350217 690ed11d   ✓

That is the same pattern MempoolSpace already uses (header_from_mempool_space_block
verifies the header hashes to the hash we asked for), except mempool.space hands
out the serialized header directly and here we assemble it.

The catch: a header costs a page of transactions

/api/v2/block/<hash> always ships up to 1000 full transactions with the header
fields; page is the only knob (pageSize is fixed at 1000 server-side), and
asking for a page past the end clamps to the last one instead of returning an
empty tx list. Measured on block 962050 (3167 txs): 1.1 MB for page=1, 161 kB
for the last page (167 txs). So the cheapest header fetch is
?page=<large> at roughly 1 kB per transaction on the last page - on average a
few hundred kB per header. (Blockbook does serve gzip - 40 kB for that 161 kB
response - but minreq doesn't decompress, so we'd pay the full size.)

That rules out backfilling and makes this a tip-following backend, same
reasoning as mempool.space: batch_header_fetch: false, run it next to a
Bitcoin Core node that supplies the active chain so new_active_headers stops
at the first header it already knows, and keep min_fork_height close to the
tip. Worth spelling out in the README like the mempool.space note does.

Other findings

• Blockbook only indexes the main chain: /api/v2/block/<stale hash> answers
  400 {"error":"Block not found"}, so no fork branches, no stale blocks from
  block(). partial_nonactive_headers: true is harmless insurance but never
  kicks in, since our own tips() only reports the active tip.
• /api/v2/block-index/<height above tip> answers 200 {"blockHash":""} rather
  than an error - needs an explicit empty check.
• /api/v2/rawtx/<txid> is not on every deployment (the instance above falls
  through to the status handler), while /api/v2/tx-specific/<txid> (bitcoind's
  getrawtransaction verbose output, including hex) is there and portable.
  Prefer the latter for the coinbase.
• Pool identification works: we can get the raw coinbase, unlike Electrum.
• Optional follow-up: Blockbook's WebSocket (/websocket) has
  subscribeNewBlock, which would fit wait_for_tip_change() and avoid polling.
  Needs a WebSocket dependency, so probably not in a first version.

Sketch

• src/backend/blockbook.rs with a Blockbook struct (info, api_url),
  HeaderFetchType::Hash, batch_header_fetch: false
• Backend::Blockbook in src/config.rs ("blockbook"), FetchError::BlockbookREST
• rpc_host = "https://<host>/api/v2", following the Esplora style
• README section + config.toml.example entry, incl. the backfill limitation

[0xB10C]

fwiw: Blockbook - blockchain indexer for Trezor Suite https://trezor.io/trezor-suite. Do not use for any other purpose.