nLockTime
and OP_CHECKLOCKTIMEVERIFY
(or CLTV) are just one side of the timelock equation. On the other side are nSequence
and OP_CHECKSEQUENCEVERIFY
, which can be used to check against relative times rather than absolute times.
⚠️ VERSION WARNING: CSV became available with Bitcoin Core 0.12.1, in spring 2016.
Every input into a transaction has an nSequence
(or if you prefer sequence
) value. It's been a prime tool for Bitcoin expansions as discussed previously in §5.2: Resending a Transaction with RBF and §8.1 Sending a Transaction with a Locktime, where it was used to signal RBF and nLockTime
, respectively. However, there's one more use for nSequence
, described by BIP 68: you can use it to create a relative timelock on a transaction.
A relative timelock is a lock that's placed on a specific input of a transaction and that's calculated in relation to the mining date of the UTXO being used in the input. For example, if a UTXO was mined at block #468260 and a transaction was created where the input for that UTXO was given an nSequence
of 100, then the new transaction could not be mined until at least block #468360.
Easy!
ℹ️ NOTE — SEQUENCE: This is the third use of the
nSequence
value in Bitcoin. AnynSequence
value without the 32nd bit set (1<<31), so 0x00000001 to 0x7ffffffff, will be interpreted as a relative timelock ifnVersion ≥ 2
(which is the default starting in Bitcoin Core 0.14.0). You should be careful to ensure that relative timelocks don't conflict with the other two uses ofnSequence
, for signallingnTimeLock
and RBF.nTimeLock
usually sets a value of 0xffffffff-1, where a relative timelock is disallowed; and RBF usually sets a value of "1", where a relative timelock is irrelevent, because it defines a timelock of 1 block.
In general, remember: with a
nVersion
value of 2, anSequence
value of 0x00000001 to 0x7fffffff allows relative timelock, RBF, andnTimeLock
; anSequence
value of 0x7fffffff to 0xffffffff-2 allows RBF andnTimeLock
; anSequence
value of 0xffffffff-1 allows onlynTimeLock
; anSequence
value of 0xffffffff allows none; andnVersion
can be set to 1 to disallow relative timelocks for any value ofnSequence
. Whew!
The format for using nSequence
to represent relative time locks is defined in BIP 68 and is slightly more complex than just inputting a number, like you did for nTimeLock
. Instead, the BIP specifications breaks up the four byte number into three parts:
- The first two bytes are used to specify a relative locktime.
- The 23rd bit is used to positively signal if the lock refers to a time rather than a blockheight.
- The 32nd bit is used to positively signal if relative timelocks are deactivated.
With that said, the construction of a block-based relative timelock is still quite easy, because the two flagged bits are set to 0
, so you just set nSequence
to a value between 1 and 0xffff (65535). The new transaction can be mined that number of blocks after the associated UTXO was mined.
You can instead set nSequence
as a relative time, where the lock lasts for 512 seconds times the value of nSequence
.
In order to do that:
- Decide how far in the future to set your relative timelock.
- Convert that to seconds.
- Divide by 512.
- Round that value up or down and set it as
nSequence
. - Set the 23rd bit to true.
To set a time 6 months n the future, you must first calculate as follows:
$ seconds=$((6*30*24*60*60))
$ nvalue=$(($seconds/512))
Then, turn it into hex:
$ hexvalue=$(printf '%x\n' $nvalue)
Finally, bitwise-or the 23rd bit into the hex value you created:
$ relativevalue=$(printf '%x\n' $((0x$hexvalue | 0x400000)))
$ echo $relativevalue
4224679
If you convert that back you'll see that 4224679 = 10000000111011010100111. The 23rd digit is set to a "1"; meanwhile the first 2 bytes, 0111011010100111, convert to 76A7 in hex or 30375 in decimal. Multiply that by 512 and you get 15.55 million seconds, which indeed is 180 days.
So you want to create a simple transaction with a relative timelock? All you have to do is issue a transaction where the nSequence
in an input is set as shown above: with the nSequence
for that input set such that the first two bytes define the timelock, the 23rd bit defines the type of timelock, and the 32nd bit is set to false.
Issue the transaction and you'll see that it can't legally be mined until enough blocks or enough time has passed beyond the time that the UTXO was mined.
Except pretty much no one does this. The BIP 68 definitions for nSequence
were incorporated into Bitcoin Core at the same time as BIP 112. which describes the CSV opcode, the nSequence
equivalent to the CLTV opcode. Just like CLTV, CSV offers increased capabilities. So, almost all usage of relative timelocks has been with the CSV opcode, not with the raw nSequence
value on its own.
Absolute Timelock | Relative Timelock | |
---|---|---|
Lock Transaction | nTimeLock | nSequence |
Lock Output | OP_CHECKLOCKTIMEVERIFY | OP_CHECKSEQUENCEVERIFY |
OP_SEQUENCEVERIFY
in Bitcoin Scripts works pretty much like OP_LOCKTIMEVERIFY
.
You might require a UTXO be held for a hundred blocks past its mining:
100 OP_CHECKSEQUENCEVERIFY
Or your might make a more complex calculation to require that a UTXO be held for six months, in which case you'll end up with a more complex number:
4224679 OP_CHECKSEQUENCEVERIFY
In this case we'll use a shorthand:
<+6Months> OP_CHECKSEQUENCEVERIFY
⚠️ WARNING: Remember that a relative timelock is a time span since the mining of the UTXO used as an input. It is not a timespan after you create the transaction. If you use a UTXO that's already been confirmed a hundred times, and you place a relative timelock of 100 blocks on it, it will be eligible for mining immediately. Relative timelocks have some very specific uses, but they probably don't apply if your only goal is to determine some set time in the future.
CSV has many of the same subtleties in usage as CLTV:
- The
nVersion
field must be set to 2 or more. - The
nSequence
field must be set to less than 0x80000000. - When CSV is run, there must be an operand on the stack that's between 0 and 0xf0000000-1.
- Both the stack operand and
nSequence
must have the same value on the 23rd bit. - The
nSequence
must be greater than or equal to the stack operand.
Just as with CLTV, when you are respending a UTXO with a CSV in its locking conditions, you must set the nSequence
to enable the transaction. You'll usually set it to the exact value in the locking script.
Just like OP_CHECKLOCKTIMEVERIFY
, OP_CHECKSEQUENCEVERIFY
includes an implicit OP_VERIFY
and leaves its arguments on the stack, requiring an OP_DROP
when you're all done.
A script that would lock funds until six months had passed following the mining of the input, and that would then require a standard P2PKH-style signature would look as follows:
<+6Months> OP_CHECKSEQUENCEVERIFY OP_DROP OP_DUP OP_HASH160 <pubKeyHash> OP_EQUALVERIFY OP_CHECKSIG
When you encode a CSV script, be careful how you encode the integer value for the relative locktime. It should be passed as a 3-byte integer, which means that you're ignoring the top byte, which could inactivate the relative locktime. Since it's an integer, be sure you convert it to little-endian.
This can be done with the integer2lehex.sh
shell script from the previous chapter.
For a relative time of 100 blocks:
$ ./integer2lehex.sh 100
Integer: 100
LE Hex: 64
Length: 1 bytes
Hexcode: 0164
Though that should be padded out to 000064
, requiring a code of 03000064
.
For a relative time of 6 months:
$ ./integer2lehex.sh 4224679
Integer: 4224679
LE Hex: a77640
Length: 3 bytes
Hexcode: 03a77640
To spend a UTXO locked with a CSV script, you must set the nSequence
of that input to a value greater than the requirement in the script, but less than the time between the UTXO and the present block. Yes, this means that you need to know the exact requirement in the locking script ... but you have a copy of the redeemScript
, so if you don't know the requirements, you deserialize it, and then set the nSequence
to the number that's shown there.
nSequence
and CSV offer an alternative to nLockTime
and CLTV where you lock a transaction based on a relative time since the input was mined, rather than basing the lock on a set time in the future. They work almost identically, other than the fact that the nSequence
value is encoded slightly differently than the nLockTime
value, with specific bits meaning specific things.
🔥 What is the power of CSV? CSV isn't just a lazy way to lock, when you don't want to calculate a time in the future. Instead, it's a totally different paradigm, a lock that you would use if it was important to create a specific minimum duration between when a transaction is mined and when its funds can be respent. The most obvious usage is (once more) for an escrow, when you want a precise time between the input of funds and their output. However, it has much more powerful possibilities in off-chain transactions, including payment channels. These applications are by definition built on transactions that are not actually put onto the blockchain, which means that if they are later put on the blockchain an enforced time-lapse can be very helpful. Hashed Timelock Contracts have been one such implementation, empowering the Lightning payment network. They're discussed in §13.3: Empowering Bitcoin with Scripts.
Advance through "Bitcoin Scripting" with Chapter Twelve: Expanding Bitcoin Scripts.