LibraryConsensus2020Design paperCorpus record
Streamlet: Textbook Streamlined Blockchains
Streamlet. Benjamin Y. Chan and Elaine Shi.
A deliberately small consensus protocol: epochs, a leader, a vote, and a rule that three adjacent certified blocks on the same chain become final. The point is to show how little machinery the last few years of BFT designs actually need.
Streamlet finalises a block when three certified blocks sit in a row on the longest certified chain. A leader proposes, a quorum certifies, and the protocol is short enough to re-derive. The replica set is known. One third faulty is still the bound.
The five-minute read
Longest certified chain
The leader extends the longest chain of blocks that already have a quorum certificate. Uncertified tips do not count as the place to build.
One vote, not three phases
The paper's streamlining is to drop the classical pre-prepare, prepare, commit theatre down to a single vote that produces a certificate. The safety argument is then carried by the three-in-a-row rule.
Three in a row
Certificates in three consecutive epochs finalise the prefix. That rule is the entire finality gadget. If a design needs a paragraph of extra cases, it is no longer the textbook version.
Known replicas
Leaders are elected from a fixed set. Streamlet does not solve membership.
One action, walked through
- An epoch begins and a leader is determined.
- The leader proposes a block on the longest certified chain.
- Replicas vote if the proposal is timely and well formed. A quorum certificate freezes that vote.
- Anyone who sees three consecutive certified blocks finalises the chain up to the first of them.
- A bad leader is skipped by the next epoch. Safety does not depend on this leader having been honest.
The argument, unpacked
A baseline, not a product
The authors wrote Streamlet so the reader can see what the previous five years of BFT had condensed to. Quoting it as a novel consensus for a public chain skips the permissioned assumption.
Partial synchrony is still there
Safety holds through asynchrony in the usual BFT way. Liveness needs a long enough synchronous period with honest leaders. A network that is never timely does not get the liveness theorem.
What has to be true
- Fewer than one third of replicas are faulty.
- The set of replicas is known and signatures identify them.
- Epochs progress. A stalled clock is outside the happy path.
- Quorum certificates are actually checked. A client that trusts a leader's word is not running Streamlet.
What happened after the paper
Streamlet is the short citation next to PBFT, Tendermint and HotStuff. Use it to test whether a new whitepaper has a new idea or a new vocabulary for certify-then-finalise.
What to check before you use the idea
- Is the voter set known in advance?
- What exactly causes finality: three certificates, or something else?
- Which faults are safety faults, and which are liveness faults?
- Does the implementation add rounds the paper deleted?
Terms
- Certified block
- A block that has gathered a quorum of votes for the proposal in its epoch.
- Three-chain rule
- The Streamlet finality rule: three adjacent certified blocks commit the prefix.
The problem the paper names
PBFT is correct and hard to re-derive. Streamlet asks for a protocol a reader can hold in their head, with the same style of quorum and a partial-synchrony assumption, without the classical message types.
What the design proposes
- Each epoch has a leader who proposes one block extending the longest certified chain.
- A block is certified when a quorum of two thirds votes for it.
- Three certified blocks in a row, with consecutive epochs, finalise the middle one's parent chain.
How the mechanism is specified
- The quorum intersection argument is the familiar one: two quorums share an honest replica if fewer than one third are faulty.
- The paper assumes a known replica set. It does not elect that set.
- Liveness needs a period of synchrony and honest leaders. Safety does not wait for that period.
What this page does not treat as proven
- A textbook protocol is not a production client. Timeouts, networking and signature aggregation are left as engineering.
- Simplicity does not remove the one-third bound.
- Citing Streamlet does not make an open membership chain into a permissioned quorum.
Why a venture studio still reads it
Read it before reading a new BFT whitepaper. If the new paper cannot say what it changed relative to 'certify, then three in a row', it may be renaming Streamlet.
This is Blockchain Lab's reading of a public design paper. It is not the paper, not a copy of it, and not an offer of tokens, equity, custody or a partnership. Later network behaviour can diverge from the text. Nothing here is investment, legal or technical advice.
Research status: Design paper. Last reviewed: 1 October 2026. This is a reading of a public paper, not investment, legal or security advice.