It should have gone out later because there literally wasn’t enough time to publish the full specs for NU7 before activation, or to properly check them against the implementations. Protocol spec: Apply the NU7 ZIPs by daira · Pull Request #1446 · zcash/zips · GitHub could only be merged on November 10, four days after activation.
The risk of a serious network failure, for example a chain fork between Zakura and Zebra, was substantially increased by that. I did in fact cross-check Zakura and Zebra against the ZIPs on the day before activation, but it probably would have been too late to do anything about it if there had been any discrepancy.
I suspect your first instinct will be to claim that this had already been checked. But I know it hadn’t been, because there were spec errors that I had to fix in that PR. Fortunately they were minor.
Why did I not do that earlier? Because I was dealing with various life stuff that was not postponable (including, for example, dealing with the some of the fallout of the governance dispute).
Please note: no-one else updated those specs, and no-one else was going to update them, despite me having done everything I could to decentralize and spread expertise. We clearly do not have the necessary redundancy and collective will among Zcash organizations to keep specifications, even just for the core consensus, up to date. That is a problem. There is an imbalance between the rush to implement new stuff and the consolidation of specs and other documentation, that will cause maintenance problems and increase security risks.
That lack of respect for the importance of specs is precisely what I’m pushing back against in this subthread. In several cases, Zcash orgs have just implemented things that break interoperability, such as fee changes or undocumented changes to the key derivation tree, without any attempt to obtain feedback from other protocol developers, or even communicate to us before deployment that a change had been made. (Often this is for changes that make other software look like it is broken, not the software that is failing to implement the spec.)
The specification and assurance backlog is also a consequence of having been overloaded with work ever since the AIpocalypse of security issues starting in March, and never having been given a humane timescale on which to recover from that, before launching into a campaign of trying to optimize everything now now now. I am not exaggerating when I say that I view this as violence to my and others’ work-life balance and relationships. I am sick of the arrogance, entitlement, and disrespect that has been shown toward long-standing protocol contributors in this period.
I don’t want to be overly dramatic, but honestly the closest analogy I can think of to how we are operating Zcash right now is the Space Shuttle programme. On each Shuttle mission, it was estimated with hindsight that the risk of a catastrophic failure ranged between 1-in-10 and 1-in-90 (for most of the programme lifetime, between 1-in-21 and 1-in-47). Ours is somewhere within that range per minor Network Upgrade. For major upgrades like the original launch, Sapling, and NU5, and emergency upgrades like NU6.2, it’s likely much higher. Tachyon will be in the same category unless we change course.
(Source: Shuttle Risk Progression Per Flight, Space Shuttle Safety and Mission Assurance Office, NASA, June 2011.)
At that failure rate you’re likely to get away with quite a few launches. On the 25th flight, the crew of the Challenger died. (I was 11, I remember vividly seeing it on the news. And I remember reading the reports, later, about why it happened and all of the missed opportunities to prevent it.) On the 113th flight, the crew of the Columbia died, and with them the future of the Shuttle programme.