Just a quick follow-up on BIP91's prospects of avoiding a BIP148 chain split, because I may have left an overly pessimistic impression -
In short: the timing isn't as dire as I suggested, BUT unless concrete progress on a plan starts taking shape, esp miner support, *the split is indeed coming.*
THE GOOD NEWS: several refinements have been noted which could get BIP91 (or splitprotection, Segwit2x, etc) deployed faster:
- The lock-in window could be shortened, eg to splitprotection's 504 blocks (3.5 days)
- Of course the 80% threshold could just be reduced, eg to splitprotection's 65%
- BIP91 nodes could start signaling on bit 1 the moment bit 4 reaches lock-in, rather than waiting another period until it "activates". (Orphaning of non-bit-1-signaling blocks would probably also have to start at or shortly after the same time [1].)
Combining these approaches, *July 26* is an approximate hard deadline for >50% of miners to be running BIP91 in order to prevent the split. So, significantly less tight than my previous June 30 (or my original, erroneous "a few days ago").
THE BAD NEWS: no one should underestimate the steps that would need to be completed by that deadline:
1. Coordinate on a solution (BIP91, splitprotection, Segwit2x, BIP148 itself, ...)
2. Implement and test it
3. Convince >50% of miners to run it [2]
4. Miners upgrade to the new software and begin signaling
In particular, #3: afaict a lot of convincing is still needed before miner support for any of these reaches anything like 50%. (With the exception of Segwit2x, but it has the additional handicap that it probably needs to include deployable hard fork code, obviously ambitious in 1.5 months.)
[2] Note that >50% need to run the *solution*, eg BIP91; old BIP141 nodes signaling segwit support do *not* count, since they won't orphan non-bit-1 blocks. The impending split isn't between nodes that support segwit vs don't, but between those that reject non-segwit-supporting blocks vs don't.