On Tue, Aug 16, 2016 at 10:52 PM, Russell O'Connor via bitcoin-dev
<bitcoin-dev@lists.linuxfoundation.org> wrote:
> I see.
>
> But is it really necessary to soft fork over this issue?  Why not just make
> it a relay rule?  Miners are already incentivized to modify transactions to
> drop excess witness data and/or prioritize (versions of) transactions based
> on their cost.  If a miner wants to mine a block with excess witness data,
> it is mostly their own loss.

Relay rules are quite fragile-- people build programs or protocols not
expecting them to be violated, without proper error handling in those
cases... and then eventually some miner rips them out because they
simply don't care about them: not enforcing them won't make their
blocks invalid.

It's my general view that we should avoid blocking things with relay
rules unless we think that someday they could be made invalid... not
necessarily that they will, but that it's plausible. Then the
elimination at the relay level is just the first exploratory step in
that direction.

One should also consider adversarial behavior by miners.  For example,
I can mine blocks with mutated witnesses with a keyed mac that chooses
the mutation. The key is shared by conspirators or customers, and now
collectively we have a propagation advantage (since we know the
mutated version before it shows up).  Not the _biggest_ concern, since
parties doing this could just create their own new transactions to
selectively propagate; but doing that would require leaving behind fee
paying public transactions, while using malleability wouldn't.