You're proposing that the total work be based on the value of the individual hashes, instead of the hash targets, is that correct?
Yes but with a little twist. To compare two chains the must include blocks up to the same level.
This could incentivize a miner who happens to find a hash with an unusually low value to refrain from immediately broadcasting it. See, for example,
this attack proposed a few years ago by casascius, which becomes viable with this proposed change.
Thanks for this particular case.
To compare two chains they must include blocks up to the same level. So although our friend could through such a block the rest miners lets
say if they where 10 blocks ahead would ask him to catch them up by providing the missing 10 blocks each one with hash value equal or smaller of existing ones.
PS. I try to think of what should happen when 2 separated block chains merge again.
I don't think that this completely eliminates casascius's attack.
Evil miner M finds a lucky block, and withholds it. During this period, evil miner M prepares double-spend attacks, and can point his mining power to his new (secret) chain.
Once good miner A publishes her new (but less lucky) block, M can take advantage of his double-spends (which now have only one confirmation), wait some short period of time and then announce his superior block.
M can therefore trick anyone who chooses to accept 1-conf transactions (which admittedly isn't all that smart, nor does M have that long to do so), plus they've also been able to mine on a secret chain while the rest of the network was wasting their time on A's shorter chain.