He also points out the need to be able to reliably spend from unconfirmed transactions and suggests that sighash_noinput can accomplish that, and would like to start a new "testnet-L" similar to testnet3 with this sighash type to test malleability mitigation. In response, Mark Friedenbach explains the soft-fork plan as having the scriptPubKey be just the 20-byte hash of the redeem script and the actual scriptSig with the redeem script and signatures contained in a separate Merkle tree committed elsewhere in the block. Meanwhile, Greg Sanders explains the hardfork variant as sticking the signatures in another parallel merkle tree and not validating them if they are not needed. Glenn Tarbox questions the lack of information on Segregated Witness, but notes that it can apparently be done with a soft fork similar to P2SH, which is seen as good news.