2

A 2015 draft BIP suggested using a "hard fork bit":

The most significant bit in nVersion is defined as the hardfork bit. Currently, blocks with this header bit setting to 1 are invalid, since BIP34 interprets nVersion as a signed number and requires it to be >=2 (with BIP66, >=3). Among the 640 bits in the block header, this is the only one which is fixed and serves no purpose, and therefore the best way to indicate the deployment of a hardfork.

https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2015-July/009576.html

If bit indexes increase from left to right, then I imagine the hard fork bit occupying index 31, marked as * below (nVersion is a 32-bit value).

0                   1                   2                   3
0|1|2|3|4|5|6|7|8|9|0|1|2|3|4|5|6|7|8|9|0|1|2|3|4|5|6|7|8|9|0|1
                                                              *

According to the proposal, setting bit 31 of nVersion should cause a block to be rejected as invalid.

BIP-9 places further constraints on the available bits, specifically, bits 0-2 must be set as [0, 0, 1]:

Blocks in the STARTED state get an nVersion whose bit position bit is set to 1. The top 3 bits of such blocks must be 001, so the range of actually possible nVersion values is [0x20000000...0x3FFFFFFF], inclusive.

https://github.com/bitcoin/bips/blob/master/bip-0009.mediawiki

In addition, BIP-9 arranges version bits in increasing order from right to left. That is, version bit 0 is nVersion bit 31, version bit 1 is nVersion bit 30, and so on.

But if this were the case, then version bit 0 would be off-limits. BIP-9 says nothing about this, and in fact BIP-68 was signaled using version bit 0:

This BIP is to be deployed by "versionbits" BIP9 using bit 0.

https://github.com/bitcoin/bips/blob/master/bip-0068.mediawiki

My questions are simple:

  1. Given my diagram above, which nVersion bit index is the "hard fork bit?"
  2. If I'm correct in my interpretation (I don't believe I am), then how could BIP-68 have been signaled without causing the blocks doing so to be rejected as invalid?

Update: the source of my confusion was thinking that nVersion bit indexing and version bit indexing ran in opposite directions. The don't. To signal readiness for a proposal using version bit 1, for example, nVersion bit 1 is set. BIP-9's rules activate when an nVersion signature is detected. This signature is bits 29-31 set as 1, 0, and 0, respectively. This leaves bit 31 unset.

Peter's answer addresses question (1) above.

As for question (2) BIP-68 signaling didn't cause block rejection because bit 0 was being set, not bit 31.

Rich Apodaca
  • 2,311
  • 2
  • 14
  • 34

1 Answers1

3

Bit 31 is sometimes referred to as the hard fork bit.

BIP34 put a constraint on version numbers to be 2 or higher. As the block version is a 32-bit signed integer, setting its highest bit results in a negative number, which would violate BIP34.

As a result, any usage of that bit results in a backward incompatible change to the rules - a hard fork - that is obvious to all participants in the system.

Pieter Wuille
  • 98,249
  • 9
  • 183
  • 287
  • That seems to match my interpretation. If so, how could BIP-68 signaling call for setting bit-31 and not result in those blocks being rejected? – Rich Apodaca Aug 09 '17 at 20:59
  • BIP68 gives meaning to bit 31 of the **nSequence field** of a **transaction**. The hardfork bit is bit 31 of the nVersion field of a block. – Pieter Wuille Aug 09 '17 at 21:04
  • Oh, I see, you're confused by bit order. Don't think of bits as having an order - it's not well defined, depends on the hardware, and is generally irrelevant. nVersion is an integer, and bit X of it is the value of `(nVersion >> X) & 1`. – Pieter Wuille Aug 09 '17 at 21:07
  • BIP-68's deployment plan calls for setting version bit 0 (`nVersion` bit 31) as a signal. I understand the functionality of BIP-68 was to assign meaning to the `nSequence` field. My question comes from the deployment mechanism for BIP-68, which requires signaling nodes to set `nVersion` bit 31. At least that's what it looks like. – Rich Apodaca Aug 09 '17 at 21:08
  • I don't understand where you read that. BIP68 was signalled for using nVersion bit 0, not bit 31. – Pieter Wuille Aug 09 '17 at 21:08
  • "This BIP is to be deployed by "versionbits" BIP9 using bit 0." (from BIP-68) AFAICT, versionbits bit 0 is bit 31 in my diagram. – Rich Apodaca Aug 09 '17 at 21:11
  • Bit 0 is bit 0, bit 31 is bit 31. I don't understand why you think there is a swap? – Pieter Wuille Aug 09 '17 at 21:12
  • Let us [continue this discussion in chat](http://chat.stackexchange.com/rooms/63578/discussion-between-pieter-wuille-and-rich-apodaca). – Pieter Wuille Aug 09 '17 at 21:12