BIP: 136 source
  Layer: Applications
  Title: Bech32 Encoded Tx Position References
  Author: Велеслав <veleslav.bips@protonmail.com>
          Jonas Schnelli <dev@jonasschnelli.ch>
          Daniel Pape <dpape@dpape.com>
  Comments-Summary: No comments yet.
  Comments-URI: https://github.com/bitcoin/bips/wiki/Comments:BIP-0136
  Status: Draft
  Type: Informational
  Created: 2017-07-09
  License: BSD-2-Clause

Introduction

Abstract

This document proposes a convenient, human usable encoding to refer to a confirmed transaction position within the Bitcoin blockchain--known as "TxRef". The primary purpose of this encoding is to allow users to refer to a confirmed transaction (and optionally, a particular outpoint index within the transaction) in a standard, reliable, and concise way.

Please note: Unlike a transaction ID, "TxID", where there is a strong cryptographic link between the ID and the actual transaction, a TxRef only provides a weak link to a particular transaction. A TxRef locates an offset within a blockchain for a transaction, that may - or may not - point to an actual transaction, which in fact may change with reorganisations. We recommend that TxRefs should be not used for positions within the blockchain having a maturity less than 100 blocks.

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.

This BIP is licensed under the 2-clause BSD license.

Motivation

Since the first version of Bitcoin, TxIDs have been a core part of the consensus protocol and are routinely used to identify individual transactions between users.

However, for many use-cases they have practical limitations:

  • TxIDs are expensive for full nodes to lookup (requiring either a linear scan of the blockchain, or an expensive TxID index).
  • TxIDs require third-party services for SPV wallets to lookup.
  • TxIDs are 64 character HEX encoded values.

It is possible to reference transactions not only by their TxID, but by their location within the blockchain itself. Rather than use the 64 character TxID, an encoding of the position coordinates can be made friendly for occasional human transcription. In this document, we propose a standard for doing this.

Examples

Block #Transaction #Outpoint #TxRefTxID
000tx1:rqqq‑qqqq‑qwtv‑vjr4a5e1e4baab89f3a32518a88c31bc87f618f76673e2cc77ab2127b7afdeda33b
17010tx1:r52q‑qqpq‑qpty‑cfgf4184fc596403b9d638783cf57adfe4c75c605f6356fbc91338530e9831e9e16
45678912341tx1:y29u‑mqjx‑ppqq‑sfp2‑tt6fb8960f70667dc9666329728a19917937896fc476dfc54a3e802e887ecb4e82

Specification

A confirmed transaction position reference, or TxRef, is a reference to a particular location within the blockchain, specified by the block height and a transaction index within the block, and optionally, an outpoint index within the transaction.

Please Note: All values in this specification are encoded in little-endian format.

TxRef Considerations

It is possible for a TxRef to reference a transaction that doesn't really exist because:

  • The specified block hasn't yet been mined.
  • The transaction index is greater than the total number of transactions included within the specified block.
  • The optional outpoint index is greater than the total outpoints contained within the transaction.

Therefore, implementers must be careful not to display TxRefs to users prematurely:

  • Applications MUST NOT display TxRefs for transactions with less than 6 confirmations.
  • Application MUST show a warning for TxRefs for transactions with less than 100 confirmations.
    • This warning SHOULD state that in the case of a large reorganisation, the TxRefs displayed may point to a different transaction, or to no transaction at all.

TxRef Format

TxRef MUST use the Bech32m[1] encoding as defined in BIP-0173 and later refined in BIP-0350. The Bech32m encoding consists of:

Human-Readable Part

The HRP can be thought of as a label. We have chosen labels to distinguish between Main, Test, and Regtest networks:

  • Mainnet: "tx".
  • Testnet: "txtest".
  • Regtest: "txrt".

Separator

The separator is the character "1".

Data Part

The data part for a TxRef consists of the transaction's block height, transaction index within the block, and optionally, an outpoint index. Specific encoding details for the data are given below.

Please note: other specifications, such as the Decentralized Identifiers spec, have implicitly encoded the information contained within the HRP elsewhere. In this case they may choose to not include the HRP as specified here.

Readability

To increase portability and readability, additional separator characters SHOULD be added to the TxRef:

  • A Colon[2] ":" added after the separator character '1'.
  • Hyphens[3] "-" added after every 4 characters beyond the colon.

Encoding

Encoding a TxRef requires 4 or 5 pieces of data: a magic code denoting which network is being used; a version number (currently always 0); the block height of the block containing the transaction; the index of the transaction within the block; and optionally, the index of the outpoint within the transaction. Only a certain number of bits are supported for each of these values, see the following table for details.

Description

Possible Data Type

# of Bits used

Values

Magic Code

Chain Namespacing Code

uint8

5

3: Mainnet
4: Mainnet with Outpoint
6: Testnet
7: Testnet with Outpoint
0: Regtest
1: Regtest with Outpoint

Version

For Future Use

uint8

1

Must be 0

Block
Height

The Block Height of the Tx

uint32

24

Block 0 to Block 16777215

Transaction
Index

The index of the Tx inside the block

uint16, uint32

15

Tx 0 to Tx 32767

Outpoint
Index

The index of the Outpoint inside the Tx

uint16, uint32

15

Outpoint 0 to Outpoint 32767

Magic Notes

The magic code provides namespacing between chains:

  • For Mainnet the magic code is: 0x3, leading to an "r" character when encoded.
  • For Mainnet with Outpoint Encoded the magic code is: 0x4, leading to a "y" character when encoded.
  • For Testnet the magic code is: 0x6, leading to an "x" character when encoded.
  • For Testnet with Outpoint Encoded the magic code is: 0x7, leading to an "8" character when encoded.
  • For Regtest the magic code is: 0x0, leading to a "q" character when encoded.
  • For Regtest with Outpoint Encoded the magic code is: 0x1, leading to a "p" character when encoded.

Encoding Example

We want to encode a TxRef that refers to Transaction #1234 of Block #456789 on the Mainnet chain. We use this data in preparation for the Bech32 encoding algorithm:

Decimal
Value

Binary
Value

# of Bits
used

Bit Indexes and Values

Magic
Code

3

00000011

5

(mc04, mc03, mc02, mc01, mc00) = (0, 0, 0, 1, 1)

Version

0

00000000

1

(v0) = (0)

Block
Height

456789

00000110
11111000
01010101

24

(bh23, bh22, bh21, bh20, bh19, bh18, bh17, bh16) = (0, 0, 0, 0, 0, 1, 1, 0)
(bh15, bh14, bh13, bh12, bh11, bh10, bh09, bh08) = (1, 1, 1, 1, 1, 0, 0, 0)
(bh07, bh06, bh05, bh04, bh03, bh02, bh01, bh00) = (0, 1, 0, 1, 0, 1, 0, 1)

Transaction
Index

1234

00000100
11010010

15

(ti14, ti13, ti12, ti11, ti10, ti09, ti08) = (0, 0, 0, 0, 1, 0, 0)
(ti07, ti06, ti05, ti04, ti03, ti02, ti01, ti00) = (1, 1, 0, 1, 0, 0, 1, 0)

As shown in the last column, we take the necessary bits of each binary value and copy them into nine unsigned chars illustrated in the next table. We only set the lower five bits of each unsigned char as the bech32 algorithm only uses those bits.

7

6

5

4

3

2

1

0

Decimal
Value

Bech32
Character

data[0]

Index

na

na

na

mc04

mc03

mc02

mc01

mc00

Value

0

0

0

0

0

0

1

1

3

r

data[1]

Index

na

na

na

bh03

bh02

bh01

bh00

v0

Value

0

0

0

0

1

0

1

0

10

2

data[2]

Index

na

na

na

bh08

bh07

bh06

bh05

bh04

Value

0

0

0

0

0

1

0

1

5

9

data[3]

Index

na

na

na

bh13

bh12

bh11

bh10

bh09

Value

0

0

0

1

1

1

0

0

28

u

data[4]

Index

na

na

na

bh18

bh17

bh16

bh15

bh14

Value

0

0

0

1

1

0

1

1

27

m

data[5]

Index

na

na

na

bh23

bh22

bh21

bh20

bh19

Value

0

0

0

0

0

0

0

0

0

q

data[6]

Index

na

na

na

ti04

ti03

ti02

ti01

ti00

Value

0

0

0

1

0

0

1

0

18

j

data[7]

Index

na

na

na

ti09

ti08

ti07

ti06

ti05

Value

0

0

0

0

0

1

1

0

6

x

data[8]

Index

na

na

na

ti14

ti13

ti12

ti11

ti10

Value

0

0

0

0

0

0

0

1

1

p

The Bech32 algorithm encodes the nine unsigned chars above and computes a checksum of those chars and encodes that as well--this gives a six character checksum (in this case, utt3p0) which is appended to the final TxRef. The final TxRef given is: tx1:r29u-mqjx-putt-3p0 and is illustrated in the following table:

TxRef character indexes and descriptions

Index0123456789101112131415161718192021
Char:tx1:r29u-mqjx-putt-3p0

Outpoint Index

Some uses of TxRef may want to refer to a specific outpoint of the transaction. In the previous example, since we did not specify the outpoint index, the TxRef tx1:r29u-mqjx-putt-3p0 implicitly references the first (index 0) outpoint of the 1234th transaction in the 456789th block in the blockchain.

If instead, for example, we want to reference the second (index 1) outpoint, we need to change the magic code from 3 to 4 and would include the following in the data to be encoded:

Decimal
Value

Binary
Value

# of Bits
used

Bit Indexes and Values

Magic
Code

4

00000100

5

(mc04, mc03, mc02, mc01, mc00) = (0, 0, 1, 0, 0)

Outpoint Index

1

00000000 00000001

15

(op14, op13, op12, op11, op10, op09, op08) = (0, 0, 0, 0, 0, 0, 0)
(op07, op06, op05, op04, op03, op02, op01, op00) = (0, 0, 0, 0, 0, 0, 0, 1)

7

6

5

4

3

2

1

0

Decimal
Value

Bech32
Character

data[0]

Index

na

na

na

mc04

mc03

mc02

mc01

mc00

Value

0

0

0

0

0

1

0

0

4

y

data[9]

Index

na

na

na

op04

op03

op02

op01

op00

Value

0

0

0

0

0

0

0

1

1

p

data[10]

Index

na

na

na

op09

op08

op07

op06

op05

Value

0

0

0

0

0

0

0

0

0

q

data[11]

Index

na

na

na

op14

op13

op12

op11

op10

Value

0

0

0

0

0

0

0

0

0

q

After Bech32 encoding all twelve unsigned chars above, we get the checksum: sfp2tt. The final TxRef given is: tx1:y29u-mqjx-ppqq-sfp2-tt and is illustrated in the following table:

TxRef character indexes and descriptions

Index012345678910111213141516171819202122232425
Char:tx1:y29u-mqjx-ppqq-sfp2-tt

Decoding

The Bech32 spec defines 32 valid characters as its "alphabet". All non-Bech32-alphabet characters present in a TxRef after the Bech32 separator character MUST be ignored/removed when parsing (except for terminating characters). We do not wish to expect the users to keep their TxRefs in good form and TxRefs may contains hyphens, colons, invisible spaces, uppercase or random characters. We expect users to copy, paste, write by-hand, write in a mix of character sets, etc. Parsers SHOULD attempt to correct for these and other common errors, reporting to the user any TxRefs that violate a proper Bech32 encoding.

As of early 2021, TxRef has been in limited use for a couple of years and it is possible that there are some TxRefs in use which were created with the original specification of Bech32 before the Bech32m refinement was codified. Due to this possibility, a TxRef parser SHOULD be able to decode both Bech32m and Bech32 encoded TxRefs. In such a case, a TxRef parser SHOULD display or somehow notify the user that they are using an obsolete TxRef and that they should upgrade it to the Bech32m version. Additionally, the parser MAY also display the Bech32m version.

Rationale

Reference implementations

C Reference Implementation (supports magic codes 0x3 and 0x6): https://github.com/jonasschnelli/bitcoin_txref_code

Go Reference Implementation (supports magic codes 0x3 and 0x6): https://github.com/kulpreet/txref

C++ Reference Implementation (supports magic codes 0x3, 0x4, 0x6, 0x7, 0x0 and 0x1): https://github.com/dcdpr/libtxref/

Java Reference Implementation (supports magic codes 0x3, 0x4, 0x6, 0x7, 0x0 and 0x1): https://github.com/dcdpr/libtxref-java/

Appendices

Test Examples

The following examples show values for various combinations on mainnet and testnet; encoding block height, transaction index, and an optional output index.

TxRef

The following list gives properly encoded mainnet TxRefs and the decoded hex values (block height, transaction index)

  • tx1:rqqq-qqqq-qwtv-vjr: (0x0, 0x0)
  • tx1:rqqq-qqll-lj68-7n2: (0x0, 0x7FFF)
  • tx1:r7ll-llqq-qats-vx9: (0xFFFFFF, 0x0)
  • tx1:r7ll-llll-lp6m-78v: (0xFFFFFF, 0x7FFF)

The following list gives properly encoded testnet TxRefs and the decoded hex values (block height, transaction index)

  • txtest1:xqqq-qqqq-qrrd-ksa: (0x0, 0x0)
  • txtest1:xqqq-qqll-lljx-y35: (0x0, 0x7FFF)
  • txtest1:x7ll-llqq-qsr3-kym: (0xFFFFFF, 0x0)
  • txtest1:x7ll-llll-lvj6-y9j: (0xFFFFFF, 0x7FFF)

The following list gives valid (sometimes strangely formatted) TxRefs and the decoded values (block height, transaction index)*

  • tx1:r29u-mqjx-putt-3p0: (456789, 1234)
  • TX1R29UMQJXPUTT3P0: (456789, 1234)
  • tx1 r29u mqjx putt 3p0: (456789, 1234)
  • tx1!r29u/mqj*x-putt^^3p0: (456789, 1234)

The following list gives invalid TxRefs and the reason for their invalidity.

  • tx1:t7ll-llll-lcq3-aj4: Magic 0xB instead of 0x3.
  • tx1:rlll-llll-lu9m-00x: Version 1 instead of 0.
  • tx1:r7ll-llll-lqfu-gss2: Valid Bech32, but ten 5 bit unsigned chars instead of nine.
  • tx1:r7ll-llll-rt5h-wz: Valid Bech32, but eight 5 bit unsigned chars instead of nine.
  • tx1:r7ll-LLLL-lp6m-78v: Invalid Bech32 due to mixed case. Would decode correctly otherwise.

TxRef with Outpoints

The following list gives properly encoded mainnet TxRefs with Outpoints and the decoded values (block height, transaction index, outpoint index)

  • tx1:yqqq-qqqq-qqqq-rvum-0c: (0x0, 0x0, 0x0)
  • tx1:yqqq-qqll-lqqq-en8x-05: (0x0, 0x7FFF, 0x0)
  • tx1:y7ll-llqq-qqqq-ggjg-w6: (0xFFFFFF, 0x0, 0x0)
  • tx1:y7ll-llll-lqqq-jhf4-wk: (0xFFFFFF, 0x7FFF, 0x0)
  • tx1:yqqq-qqqq-qpqq-pw4v-kq: (0x0, 0x0, 0x1)
  • tx1:yqqq-qqll-lpqq-m3w3-kv: (0x0, 0x7FFF, 0x1)
  • tx1:y7ll-llqq-qpqq-22ml-hz: (0xFFFFFF, 0x0, 0x1)
  • tx1:y7ll-llll-lpqq-s4qz-hw: (0xFFFFFF, 0x7FFF, 0x1)
  • tx1:y29u-mqjx-ppqq-sfp2-tt: (456789, 1234, 1)

The following list gives properly encoded testnet TxRefs with Outpoints and the decoded values (block height, transaction index, outpoint index)

  • txtest1:8qqq-qqqq-qqqq-d5ns-vl: (0x0, 0x0, 0x0)
  • txtest1:8qqq-qqll-lqqq-htgd-vn: (0x0, 0x7FFF, 0x0)
  • txtest1:87ll-llqq-qqqq-xsar-da: (0xFFFFFF, 0x0, 0x0)
  • txtest1:87ll-llll-lqqq-u0x7-d3: (0xFFFFFF, 0x7FFF, 0x0)
  • txtest1:8qqq-qqqq-qpqq-0k68-48: (0x0, 0x0, 0x1)
  • txtest1:8qqq-qqll-lpqq-4fp6-4t: (0x0, 0x7FFF, 0x1)
  • txtest1:87ll-llqq-qpqq-yj55-59: (0xFFFFFF, 0x0, 0x1)
  • txtest1:87ll-llll-lpqq-7d0f-5f: (0xFFFFFF, 0x7FFF, 0x1)
  • txtest1:829u-mqjx-ppqq-73wp-gv: (456789, 1234, 1)

TxRef Payload Value Choices:

Some calculations showing why we chose these particular bit-length of the block height and transaction index.

Block Height Value:

24 bits: value can be between 0, and 0xFFFFFF (16777216 blocks).

  • In early April, 2021, there have been 677700 blocks
  • There are roughly (365 days * 24 hours * 6 blocks / hour) = 52560 blocks every year, implying about (16777216 - 677700) / 52560 = 306 more years of addressable blocks.
  • Some time before year 2327 this specification should be extended.

Tx Position Value:

15 bits: value can be between 0x0, and 0x7FFF (32768 transactions).

  • The realistic smallest Tx is 83 Bytes for maximum 12047 tx in a block.
    • 4B version + 1B tx_in count + 36B previous_output + 1B script length + 0B signature script + 4B sequence + 1B tx_out count + 8B amount + 1B script length + 23B pubkey script + 4B lock_time = 83B
  • The extreme smallest Tx is 60 Bytes for maximum 16665 tx in a block.
    • 4B version + 1B tx_in count + 36B previous_output + 1B script length + 0B signature script + 4B sequence + 1B tx_out count + 8B amount + 1B script length + 0B pubkey script + 4B lock_time = 60B

Acknowledgements

Special Thanks to Pieter Wuille and Greg Maxwell for Bech32, a wonderful user-facing data encoding.

[1] Why use Bech32 Encoding for Confirmed Transaction References? The error detection and correction properties of this encoding format make it very attractive. We expect that it will be reasonable for software to correct a maximum of two characters; however, we haven’t specified this yet.

[2] Why add a colon here? This allows it to conform better with W3C URN/URL standards.

[3] Why hyphens within the TxRef? As TxRefs are short, we expect that they will be quoted via voice or written by hand. The inclusion of hyphens every 4 characters breaks up the string and means people don't lose their place so easily.