Very cool work.
I don’t mean to be discouraging, but isn’t this more or less what we spent a bunch of time on assumeutxo to accomplish?
IBD booster requires (as far as I can tell) essentially trusted metadata – the hints – that were produced using some other, already synced node. It also requires modifications to very sensitive parts of the validation code in order to make use of the hints.
Assumeutxo snapshots, on the other hand, aren’t trusted in the same way that hints seem to be given that the snapshots are committed to with hashes in the source code.
IBD booster provides a ~2.24x speedup – which again is very cool – but assumeutxo provides a higher order of speedup (sync in an hour or two) given that we’re truncating the IBD process to the history between the snapshot base and the current network tip.
This is very interesting research, and I don’t mean to discourage. But I’m just wondering if it isn’t worth instead investing more effort to “deliver” assumeutxo in terms of making it usable with the GUI etc. rather than inventing and implementing yet another way to produce exogenous metadata, distribute it safely, and use it to speed up IBD.