MIME-Version: 1.0
Sender: gmaxwell@gmail.com
In-Reply-To: <343A3542-3103-42E9-95B7-640DFE958FFA@gmail.com>
References: <7E4FA664-BBAF-421F-8C37-D7CE3AA5310A@gmail.com>
	<F87D7069-0FDC-4572-B02B-398A2A455935@gmail.com>
	<CAAS2fgT716PiP0ucoASxryM9y+s9H2z06Z0ToaP1xT3BozAtNw@mail.gmail.com>
	<CADZtCSguto2z6Z9CykymxnCokqo1G=sW0Ov0ht+KcD+KMnYyow@mail.gmail.com>
	<CAO3Pvs-YDzfRqmyJ85wTH0ciccjCvkm5stGyP_tVGGna=PMv3A@mail.gmail.com>
	<CAO3Pvs9p5COiS_7Jbj1r2iAKTEdXUcnVTRzL27c3=CeuB9WDTQ@mail.gmail.com>
	<CAAS2fgSyVi0d_ixp-auRPPzPfFeffN=hsWhWT5=EzDO3O+Ue1g@mail.gmail.com>
	<CAO3Pvs_0qCZbRCfL8EJw6gzWjZeXWcJrtg27g_SJ7+PkYTHg6A@mail.gmail.com>
	<CAAS2fgTs+aKyiL8Kg_AZk=Mdh6896MEg=KHa6ANAZO7unsGEsg@mail.gmail.com>
	<CADZtCShyYbgKk2zsKzQniqDw--XKfYWTk3Hk3o50V=MgT6zeuQ@mail.gmail.com>
	<20180602124157.744x7j4u7dqtaa43@email>
	<343A3542-3103-42E9-95B7-640DFE958FFA@gmail.com>
From: Gregory Maxwell <greg@xiph.org>
Date: Sun, 3 Jun 2018 00:28:33 +0000
Message-ID: <CAAS2fgQDdJpzPR9Ve81hhyqU+MO7Ryy125fzK-iv=sfwwORDCw@mail.gmail.com>
To: Tamas Blummer <tamas.blummer@gmail.com>, 
	Bitcoin Protocol Discussion <bitcoin-dev@lists.linuxfoundation.org>
Content-Type: text/plain; charset="UTF-8"
Subject: Re: [bitcoin-dev] BIP 158 Flexibility and Filter Size
Precedence: list

On Sat, Jun 2, 2018 at 10:02 PM, Tamas Blummer via bitcoin-dev
<bitcoin-dev@lists.linuxfoundation.org> wrote:
> Years of experience implementing wallets with BIP 37

pretty much us that all these filter things are a total waste of time.
BIP37 use is nearly dead on the network-- monitor your own nodes to
see the actual use of the filters: it's very low.  I see under average
of 1 peer per day using it.

Moreover the primary complaint from users about BIP37 vs the
alternatives they're choosing over it (electrum and web wallets) is
that the sync time is too long-- something BIP158 doesn't improve.

So if we were going to go based on history we wouldn't bother with on
P2P at all.   But I think the history's lesson here may mostly be an
accident, and that the the non-use of BIP37 is  due more to the low
quality and/or abandoned status of most BIP37 implementing software,
rather than a fundamental lack of utility.   Though maybe we do find
out that once someone bothers implementing a PIR based scanning
mechanism (as electrum has talked about on and off for a while now)
we'll lose another advantage.

BIP37 also got a number of things wrong-- what went into the filters
was a big element in that (causing massive pollution of matches due to
useless data), along with privacy etc.  This kind of approach will
have the best chances if it doesn't repeat the mistakes... but also
it'll have the best chances if it has good security, and getting SPV-
equivalent security will require committing the filters, but
committing them is a big step because then the behaviour becomes
consensus normative-- it's worth spending a few months of extra
iteration getting the design as good as possible before doing that
(which is what we've been seeing lately).