To: Mark Friedenbach <mark@friedenbach.org>
References: <53A587C3-DAC1-4055-875F-96B61717ACE6@xbt.hk>
	<081A517B-B730-43AB-9D4E-4F696EFD91A3@friedenbach.org>
	<56ca1248-6427-46f7-1645-84349cc8facc@mattcorallo.com>
	<44B74F02-D3D6-47B9-976E-A72042E5C84B@friedenbach.org>
From: Matt Corallo <lf-lists@mattcorallo.com>
Message-ID: <2c78e653-862d-afd3-0ae7-ebbbc9eec770@mattcorallo.com>
Date: Mon, 27 Nov 2017 16:33:37 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
	Thunderbird/52.3.0
MIME-Version: 1.0
In-Reply-To: <44B74F02-D3D6-47B9-976E-A72042E5C84B@friedenbach.org>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Cc: Bitcoin Protocol Discussion <bitcoin-dev@lists.linuxfoundation.org>
Subject: Re: [bitcoin-dev] Making OP_CODESEPARATOR and FindAndDelete in
 non-segwit scripts non-standard
Precedence: list

Indeed, the PR in question does *not* change the semantics of
OP_CODESEPARATOR within SegWit redeemScripts, where it is still allowed
(and Nicolas Dorier pointed out that he was using it in TumbleBit), so
there are still ways to use it, but only in places, like SegWit, where
the potential validation complexity blowup is massively reduced.

I am not sure that OP_CODESEPARATOR is entirely useless in pre-SegWit
scripts (I believe Nicolas' construction may still be relevant
pre-SegWit), though I strongly believe FindAndDelete is.

I don't think CODESEPARATOR rises to the threshold of it being "widely
known to be useless", but certainly the historical use of it (to
separate the scriptSig and the scriptPubKey in the scriptCode, which was
run as a single concatenated thing in the original design is no longer
relevant). FindAndDelete is equally irrelevant if not significantly more
irrelevant.

Matt

On 11/27/17 16:06, Mark Friedenbach wrote:
> It is relevant to note that BIP 117 makes an insecure form of
> CODESEPARATOR delegation possible, which could be made secure if some
> sort of CHECKSIGFROMSTACK opcode is added at a later point in time. It
> is not IMHO a very elegant way to achieve delegation, however, so I hope
> that one way or another this could be resolved quickly so it doesn’t
> hold up either one of those valuable additions.
> 
> I have no objections to making them nonstandard, or even to make them
> invalid if someone with a better grasp of history can attest that
> CODESEPARATOR was known to be entirely useless before the introduction
> of P2SH—not the same as saying it was useless, but that it was widely
> known to not accomplish what a early-days script author might think it
> was doing—and the UTXO set contains no scriptPubKeys making use of the
> opcode, even from the early days. Although a small handful could be
> special cased, if they exist.
> 
>> On Nov 27, 2017, at 8:33 AM, Matt Corallo <lf-lists@mattcorallo.com
>> <mailto:lf-lists@mattcorallo.com>> wrote:
>>
>> I strongly disagree here - we don't only soft-fork out transactions that
>> are "fundamentally insecure", that would be significantly too
>> restrictive. We have generally been willing to soft-fork out things
>> which clearly fall outside of best-practices, especially rather
>> "useless" fields in the protocol eg soft-forking behavior into OP_NOPs,
>> soft-forking behavior into nSequence, etc.
>>
>> As a part of setting clear best-practices, making things non-standard is
>> the obvious step, though there has been active discussion of
>> soft-forking out FindAndDelete and OP_CODESEPARATOR for years now. I
>> obviously do not claim that we should be proposing a soft-fork to
>> blacklist FindAndDelete and OP_CODESEPARATOR usage any time soon, and
>> assume that it would take at least a year or three from when it was made
>> non-standard to when a soft-fork to finally remove them was proposed.
>> This should be more than sufficient time for folks using such weird (and
>> largely useless) parts of the protocol to object, which should be
>> sufficient to reconsider such a soft-fork.
>>
>> Independently, making them non-standard is a good change on its own, and
>> if nothing else should better inform discussion about the possibility of
>> anyone using these things.
>>
>> Matt
>>
>> On 11/15/17 14:54, Mark Friedenbach via bitcoin-dev wrote:
>>> As good of an idea as it may or may not be to remove this feature from
>>> the code base, actually doing so would be crossing a boundary that we
>>> have not previously been willing to do except under extraordinary
>>> duress. The nature of bitcoin is such that we do not know and cannot
>>> know what transactions exist out there pre-signed and making use of
>>> these features.
>>>
>>> It may be a good idea to make these features non standard to further
>>> discourage their use, but I object to doing so with the justification of
>>> eventually disabling them for all transactions. Taking that step has the
>>> potential of destroying value and is something that we have only done in
>>> the past either because we didn’t understand forks and best practices
>>> very well, or because the features (now disabled) were fundamentally
>>> insecure and resulted in other people’s coins being vulnerable. This
>>> latter concern does not apply here as far as I’m aware.
>>>
>>> On Nov 15, 2017, at 8:02 AM, Johnson Lau via bitcoin-dev
>>> <bitcoin-dev@lists.linuxfoundation.org
>>> <mailto:bitcoin-dev@lists.linuxfoundation.org>
>>> <mailto:bitcoin-dev@lists.linuxfoundation.org>> wrote:
>>>
>>>> In https://github.com/bitcoin/bitcoin/pull/11423 I propose to
>>>> make OP_CODESEPARATOR and FindAndDelete in non-segwit scripts
>>>> non-standard
>>>>
>>>> I think FindAndDelete() is one of the most useless and complicated
>>>> functions in the script language. It is omitted from segwit (BIP143),
>>>> but we still need to support it in non-segwit scripts. Actually,
>>>> FindAndDelete() would only be triggered in some weird edge cases like
>>>> using out-of-range SIGHASH_SINGLE.
>>>>
>>>> Non-segwit scripts also use a FindAndDelete()-like function to remove
>>>> OP_CODESEPARATOR from scriptCode. Note that in BIP143, only executed
>>>> OP_CODESEPARATOR are removed so it doesn’t have the
>>>> FindAndDelete()-like function. OP_CODESEPARATOR in segwit scripts are
>>>> useful for Tumblebit so it is not disabled in this proposal
>>>>
>>>> By disabling both, it guarantees that scriptCode serialized inside
>>>> SignatureHash() must be constant
>>>>
>>>> If we use a softfork to remove FindAndDelete() and OP_CODESEPARATOR
>>>> from non-segwit scripts, we could completely remove FindAndDelete()
>>>> from the consensus code later by whitelisting all blocks before the
>>>> softfork block. The first step is to make them non-standard in the
>>>> next release.
>>>>
>>>>
>>>>  
>>>> _______________________________________________
>>>> bitcoin-dev mailing list
>>>> bitcoin-dev@lists.linuxfoundation.org
>>>> <mailto:bitcoin-dev@lists.linuxfoundation.org>
>>>> <mailto:bitcoin-dev@lists.linuxfoundation.org>
>>>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>>>
>>>
>>> _______________________________________________
>>> bitcoin-dev mailing list
>>> bitcoin-dev@lists.linuxfoundation.org
>>> <mailto:bitcoin-dev@lists.linuxfoundation.org>
>>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>