Question:

Why are miners allowed to change a tx id?

Benjamin: 5 weeks ago

BIP 62 (https://github.com/bitcoin/bips/blob/master/bip-0062.mediawiki) proposed to make transaction malleability impossible.

Why was this retracted and why are miners allowed to change a transaction id?

Also, what does it mean when someone says signatures cannot sign themselves? Is this implying we not only need a signature for the transaction, we need another signature for the transaction + signature?

Answer:
James: 5 weeks ago

BIP 62 proposed to make transaction malleability impossible.

No it didn't. It would have outlawed a limited number of known and addressable forms of malleability.

In particular, it did not prevent multisig transaction participants from just creating a new signature, malleating the transaction without invalidation. This type of malleability is trivial to perform, and breaks nearly every smart contract construction imaginable.

BIP62 was retracted because while it addressed some forms of malleability, it didn't and couldn't fix an equally bad one.

Instead, Bitcoin adopted BIP141 (Segregated Witness), which does not remove malleability, but makes it harmless. The transaction signatures just don't contribute to the txids anymore. Thus, while it is still possible to malleate transactions, doing so does not invalidate dependent transactions (as they refer to previous transactions using the txid). Other than BIP62, this solution applies to multisignatures as well, and also against all forms of non-discovered malleability.