the way ConsumePrelude works is that it reads one byte at a time, and every time it reads a byte, it checks whether the end marker is in the entire string. This is O(N^2) at best... and destroys performance with non-trivial sizes of metadata.
we don't need preludes anymore, since zfs user props are no longer in use.
the way
ConsumePrelude
works is that it reads one byte at a time, and every time it reads a byte, it checks whether the end marker is in the entire string. This is O(N^2) at best... and destroys performance with non-trivial sizes of metadata.we don't need preludes anymore, since zfs user props are no longer in use.