From: Luke Dashjr <luke@dashjr.org>
To: bitcoin-dev@lists.linuxfoundation.org,
 Erik Aronesty <erik@q32.com>
Date: Tue, 21 Jun 2016 20:44:37 +0000
User-Agent: KMail/1.13.7 (Linux/4.1.18-gentoo; KDE/4.14.16; x86_64; ; )
References: <CAJowKg+zYtUnHv+ea--srehVa5K46sjpWbHVcVGRY5x0w5XRTQ@mail.gmail.com>
In-Reply-To: <CAJowKg+zYtUnHv+ea--srehVa5K46sjpWbHVcVGRY5x0w5XRTQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="iso-8859-15"
Content-Transfer-Encoding: 7bit
Message-Id: <201606212044.38931.luke@dashjr.org>
Subject: Re: [bitcoin-dev] Even more proposed BIP extensions to BIP 0070
Precedence: list

On Monday, June 20, 2016 5:33:32 PM Erik Aronesty via bitcoin-dev wrote:
> BIP 0070 has been a a moderate success, however, IMO:
> 
> - protocol buffers are inappropriate since ease of use and extensibility is
> desired over the minor gains of efficiency in this protocol.  Not too late
> to support JSON messages as the standard going forward

IMO JSON is too prone to gratuitous inefficiency (both at network and CPU 
level), parser bugs, etc. Even the best C implementation (jansson) has serious 
issues with Number handling.

A few years ago, I looked into binary alternatives to JSON and concluded they 
all had problems, while it seems more than reasonable to do even dynamic 
parsing of protobuf messages. So to conclude, I prefer to stick to protobuf 
unless a clearly superior protocol turns up.

> - problematic reliance on merchant-supplied https (X509) as the sole form
> of mechant identification.   alternate schemes (dnssec/netki), pgp and
> possibly keybase seem like good ideas.   personally, i like keybase, since
> there is no reliance on the existing domain-name system (you can sell with
> a github id, for example)

X509 is entrenched, so it should remain supported. PGP might make sense for 
people already using it (it provides no real security for un-WoT-networked 
users), but unforunately, few people use it. Correct me if I'm wrong, but IIRC 
Keybase uses blockchain spam, so definitely not something to be encouraged if 
so. Namecoin seems like a more than reasonable decentralised solution, but 
will probably take some real work to implement (not that this is avoidable for 
a general-usage decentralised solution).

> - missing an optional client supplied identification

What do you mean by this? There's the memo field at least.

> - lack of basic subscription support
> 
> *Proposed for subscriptions:*
> 
> - BIP0047 payment codes are recommended instead of wallet addresses when
> establishing subscriptions.  Or, merchants can specify replacement
> addresses in ACK/NACK responses.   UI confirms are *required *when there
> are no replacement addresses or payment codes used.

I'd discourage anything using BIP 47 due to its serious design flaws.
No reason a regular BIP 32 pub seed can't be used instead.

What do you mean by "replacement addresses" and "UI confirms" here?

> - Wallets must confirm and store subscriptions, and are responsible for
> initiating them at the specified interval.
> 
> - Intervals can *only *be from a preset list: weekly, biweekly, or 1,
> 2,3,4,6 or 12 months.   Intervals missed by more than 3 days cause
> suspension until the user re-verifies.

Disagree with hard-coding intervals, or mandating specific policies from the 
service providers.

> - Wallets *may *optionally ask the user whether they want to be notified
> and confirm every interval - or not.   Wallets that do not ask *must
> *notify before initiating each payment.   Interval confirmations should
> begin at *least *1 day in advance of the next payment.

This is wallet policy, but maybe makes sense as a "best practices" BIP.

> *Proposed in general:*
> - JSON should be used instead of protocol buffers going forward.  Easier to
> use, explain extend.
> 
> - "Extendible" URI-like scheme to support multi-mode identity mechanisms on
> both payment and subscription requests.   Support for keybase://, netki://
> and others as alternates to https://.
> 
> - Support for client as well as merchant multi-mode verification
> 
> - Ideally, the identity verification URI scheme is somewhat
> orthogonal/independent of the payment request itself
> 
> Question:
> 
> Should this be a new BIP?  I know netki's BIP75 is out there - but I think
> it's too specific and too reliant on the domain name system.
>
> Maybe an identity-protocol-agnostic BIP + solid implementation of a couple
> major protocols without any mention of payment URI's ... just a way of
> sending and receiving identity verified messages in general?
> 
> I would be happy to implement plugins for identity protocols, if anyone
> thinks this is a good idea.
> 
> Does anyone think https:// or keybase, or PGP or netki all by themselves,
> is enough - or is it always better to have an extensible protocol?
> 
> - Erik Aronesty