Currently only directly invalid blocks are included in this dataset, with descendants of invalid blocks being held out.
Strictly, such blocks are also invalid by their ancestry, so they should be included. Specifically, Bitcoin Core treats these as invalid, returning bad-prevblk for a new header whose parent is already marked invalid.
I think we need to expand the remit of classification so that they indeed are included, but thought to pause before I pulled the trigger on this. I guess this is just another mode of failure, albeit one that is completely externally contextual as opposed to some failure in block template construction.
Currently only directly invalid blocks are included in this dataset, with descendants of invalid blocks being held out.
Strictly, such blocks are also invalid by their ancestry, so they should be included. Specifically, Bitcoin Core treats these as invalid, returning
bad-prevblkfor a new header whose parent is already marked invalid.I think we need to expand the remit of classification so that they indeed are included, but thought to pause before I pulled the trigger on this. I guess this is just another mode of failure, albeit one that is completely externally contextual as opposed to some failure in block template construction.