← Index
ARTEL 21: Block Layout
ARTEL 21 / Articles

The Block Layout, Before and After

Eighty bytes of header, a variable number of transactions, and four consensus changes that have moved what goes where.

A Bitcoin block is a serialized byte string. Its layout — the order, size, and meaning of every field — is defined by consensus. The 80-byte header has not changed since block 0. Everything underneath it has: the coinbase grew a height field, transactions grew a witness section, and the accounting unit moved from bytes to weight. This article shows the four eras side by side at a single 4 MB scale, then walks the key changes in prose.

All four eras, one scale

Every bar is drawn on the same 4,000,000-byte axis. The pre-SegWit eras fill only the first quarter (1 MB cap). The post-SegWit eras use the full width. The header is 80 B in every era — the same gold tab at the far left. What changes is what follows.

Block layout — 0 to 4,000,000 B
all eras at 4 MB scale
Coinbase across eras
tx 0, annotated
Transaction across eras
representative tx, annotated

What changed, and when

The header never changed. From genesis to the present, the first 80 bytes of every block contain the same six fixed-width fields: version (4 B), previous block hash (32 B), merkle root (32 B), timestamp (4 B), bits (4 B), and nonce (4 B). The merkle root commits to transaction identifiers; under SegWit it commits to wtxids instead of txids, but the field itself is byte-identical.

BIP-34 (March 2013, block 227,836) added the block height as the first push in the coinbase scriptsig. Before this, the coinbase input could carry arbitrary data. After BIP-34, it must begin with a CompactSize varint encoding the height — one byte up to block 65,535, three bytes above. This pins every coinbase to its height and prevents duplicate coinbases across competing chains. Regular transactions were unaffected.

BIP-141 (August 2017, block 481,824) introduced Segregated Witness. Signature data was moved out of each input's scriptsig and into a new witness section that follows the legacy portion of the transaction. The coinbase gained a second output: a 38-byte OP_RETURN witness commitment that anchors the wtxid merkle root. A 32-byte witness nonce was added to the coinbase witness. The accounting unit changed from raw bytes to weight, defined as 4 × stripped + witness ≤ 4,000,000. The effective ceiling became roughly 1 MB of legacy data plus up to 3 MB of witness data.

BIP-341 (November 2021, block 709,635) introduced Taproot. The block header, coinbase, and weight accounting remained identical to SegWit. What changed was inside the witness: a new v1 output type (P2TR, 32-byte x-only key), 64-byte Schnorr signatures replacing ~71–72 B DER ECDSA signatures, and MAST — the ability to commit to a tree of alternative spending conditions and reveal only the one taken. An optional annex, prefixed with 0x50, can follow any witness and is readable by scripts.

Blocksize as universe, witness as unequal bits

The redshift model in Bitcoin Redshift treats the 262.5 TB addressable space as a closed container. What the layout history reveals is that the container has been resized twice — and that its internal geometry is non-uniform.

Block size is the universe boundary. In the Original and BIP-34 eras the cap was 1 MB. Under that ceiling, reaching the 262.5 TB memory wall takes ~5,000 years at full utilization. BIP-141 expanded the effective ceiling to ~4 MB by introducing weight accounting, widening the aperture through which information enters the same fixed container. A 4 MB universe fills proportionally faster — or, equivalently, has less time in Expansion mode before entering Bounded Dynamics. The block size is not merely a throughput limit; it is the aperture of the closed container.

Weight breaks bit equality. Under BIP-141 one byte of stripped data costs four weight units; one byte of witness data costs one weight unit. A bit in the witness is four times cheaper than a bit in the stripped section. This means 1 sat = 1 bit is only true in the most abstract sense. In practice, the bits are not fungible. A Schnorr signature (64 B witness) consumes less weight than a single P2PKH output (34 B stripped). The block is a weighted space, not a flat memory space.

The gradient shapes structure. The 4:1 discount creates a thermodynamic gradient: validation logic (witness) is cheap to include, while UTXO state (stripped) is expensive. This is intentional. It incentivizes keeping the permanent set small while allowing complex scripts to be verified and discarded. From the redshift perspective, the witness discount accelerates the rate at which bits are set into structure — blocks are larger, so the chain fills the container faster — while shaping where that structure lives. The trace grows faster; the permanent state grows slower. The container is weighted.

The four eras, side by side

The block header has not changed. The coinbase grew twice — once for the height, once for the witness commitment. The transaction is the only place where witness data, and later Taproot's v1, Schnorr, MAST, and annex, live. Weight replaced bytes as the accounting unit, and the effective block ceiling climbed from 1 MB to roughly 4 MB.

Dimension Original + BIP-34 + BIP-141 + BIP-341
Header 80 B
6 fields
80 B
6 fields
80 B
6 fields
80 B
6 fields
Coinbase version + 1 input (no height) + 1 output + locktime + varint(height) at start of scriptsig + 38 B witness commitment output
+ 32 B witness nonce
+ 38 B witness commitment output
+ 32 B witness nonce
Tx version + ins + outs + locktime unchanged + witness section per input + v1 (Schnorr, MAST, annex)
Witness absent absent v0 — P2WPKH, P2WSH v0 + v1 (P2TR)
Block limit 1,000,000 B 1,000,000 B 4,000,000 weight 4,000,000 weight
Effective ceiling ~1 MB ~1 MB ~1 MB stripped + ~3 MB witness ~1 MB stripped + ~3 MB witness
Activated block 0
Jan 2009
block 227,836
Mar 2013
block 481,824
Aug 2017
block 709,635
Nov 2021

BIPs cited

Related: Bitcoin Memory — a theoretical model treating the sat supply as 262.5 TB of addressable RAM, consumed block by block.

Related: Bitcoin Redshift — eight observable signals of stretching, cooling, and dilution in the Bitcoin system, with the memory saturation model as one boundary condition.