Re: [bitcoin-dev] BIP 174 thoughts

From: Andrew Chow <achow101-lists@achow101•com>
To: Andrea <a.raspitzu@protonmail•com>,
	Bitcoin Protocol Discussion
	<bitcoin-dev@lists•linuxfoundation.org>
Subject: Re: [bitcoin-dev] BIP 174 thoughts
Date: Sun, 24 Jun 2018 04:28:26 -0400	[thread overview]
Message-ID: <4BLLfQQ5BFO3Z2E2NagJ5trtBmdr6if2KSR9gWpYQY2xKu6THdvk0LJbkRxr8Yie2HA17KOZIM2ljupV_H8cfVkGFFcRjOrA0b13KG9ciF4=@achow101.com> (raw)
In-Reply-To: <HWG4r9Lgi0mPONMhHwAXqnGPYeqeAvarQBUlqaRa-iZeysawpb2CN76M0ywrxbhLGJirwWViKIJqwadJcjYPRdYff4ISkSYXAO4a0SWBdVA=@protonmail.com>

I disagree with the idea that global types can be removed. Firstly, it
is less of a breaking change to leave it there than to remove it
entirely. Secondly, there may be a point in the future where global
types would be useful/necessary. By having it still be there, we allow
for future extensibility.

Andrew

On 06/24/2018 01:19 AM, Andrea wrote:
> Hi, 
>
> I think in the revised spec we can remove completely the "global types" as a map or even as typed record. Since there is only one type (the transaction) and it's compulsory to have one (and only one) we could just drop the definition of global type and the key associated with it, simply after the header + separator there must be a transaction. Having read all the discussion i also agree having per-input key derivation and per-output data is a lot more handy for signers, no special feeling regarding the encoding.Looking forward for the test vectors and the new spec.
>
> Cheers, Andrea.
>
> ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
>
> On June 23, 2018 10:33 PM, Andrew Chow via bitcoin-dev <bitcoin-dev@lists•linuxfoundation.org> wrote:
>
>> 
>>
>> On 06/23/2018 10:00 AM, William Casarin wrote:
>>
>>> Since we're still considering the encoding, I wonder if it would be a
>>>
>>> good idea to have a human-readible part like lightning invoices[1]?
>> I don't think that is necessary.
>>
>>> Then perhaps you could drop the magic code as well?
>> The magic is still necessary for the binary format in order to prevent
>>
>> normal transaction deserializers from accidentally deserializing a psbt.
>>
>>> Also we could do a base encoding that excludes + and / characters, such
>>>
>>> as base62 (gmp-style). It's easier to copy/paste (double clicking a
>>>
>>> string stops at / or + in base64 encodings).
>> While that would be ideal, I think it is better to use an encoding that
>>
>> most wallets already support. Most wallets already have Base64 decoding
>>
>> available so that they can decode signed messages which also use Base64
>>
>> encoding. I think it is unnecessary to introduce another encoding.
>>
>> On 06/23/2018 11:27 AM, Peter D. Gray wrote:
>>
>>> Personally, I don't think you should spec an encoding. It should be binary only and hex for developers and JSON interfaces. My thinking is that PSBT's are not user-visible things. They have a short lifetime and are nothing something that is "stored" or referenced much later. Hex is good enough and has no downsides as an excoding except for density.
>> I think what will end up happening though is that, at least in the
>>
>> beginning, PSBTs will primarily be strings that people end up copy and
>>
>> pasting. Since a PSBT can get pretty large, the strings are rather
>>
>> cumbersome to move around, especially as hex. At least with Base64 the
>>
>> strings will be smaller.
>>
>>> On the other hand, suggesting a filename extension (probably .PSBT?) and a mime-type to match, are helpful since wallets and such will want to register with their operating systems to become handlers of those mimetypes. Really that's a lot more important for interoperability at this point, than an encoding.
>> Agreed. I will include those in the BIP.
>>
>>> Looking forward to test vectors, and I might have more to say once my code can handle them (again).
>>>
>>> Feedback on the BIP as it stands now:
>>>
>>> -   Appendix A needs an update, and I suggest defining symbols (PK_PARTIAL_SIG == 0x02) for the numeric key values. This helps implementers as we don't all define our own symbols and/or use numeric constants in our code.
>> Okay.
>>
>>> -   Those tables are just not working. Might want to reformat as descriptive lists, point form, or generally anything else... sorry.
>> I will try my best to fix that. Mediawiki sucks...
>>
>> Andrew
>>
>> bitcoin-dev mailing list
>>
>> bitcoin-dev@lists•linuxfoundation.org
>>
>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev