MIME-Version: 1.0
Sender: gmaxwell@gmail.com
In-Reply-To: <CABm2gDqmOXGZ_qxyNjFYg8qTRu4Tmo7V+zYZCt0j5RekUNVkWw@mail.gmail.com>
References: <CAEgR2PGCZ=F85yjAbZgC6NtzhpdgBL3n4M2jowN12wJ7x-Ai1A@mail.gmail.com>
	<CAEgR2PGrxDQE0k8WX4XXz9GN-RAL6JB51ST9Hdz=ba36gRCa6A@mail.gmail.com>
	<CAEgR2PFjt=ihzRBhNXbHTAJz1R+3vz8o-zRZkDA3iBo39x9cTQ@mail.gmail.com>
	<CAEgR2PFfSjJjkTYq+DAmTzmkHPxqhn6fUDoXTzrRebz+OoUgqw@mail.gmail.com>
	<CAEgR2PG5ZueHKDXbsPDEjQG7xAYBa_JAtPZo9n1V2=STC1srpA@mail.gmail.com>
	<CAEgR2PGPQ1e9SmoWOS3V+N9v+OWiM4g3nPN3d9urc+DfkWEJ7A@mail.gmail.com>
	<CAEgR2PEKkHH6+Sh8cQGF83-s1tpwQZgd0fiuNz_xyWu0mUPfCA@mail.gmail.com>
	<CAEgR2PEyWFO1RFohVEpcb-M7aM-8xjCFvDPeJPD4zF4yTCyZ0A@mail.gmail.com>
	<CAEgR2PGrf+4pQRyNC_xKVEKXimKTWveGK9q6YJeZkG0_r=8tkg@mail.gmail.com>
	<5F7A4F74-B108-4E30-A3F4-4125BBD0F819@friedenbach.org>
	<CABm2gDqXXvNCZ7EyKuwudB5J0YDX7hNnXHPZNxTO0_JsM+yNHg@mail.gmail.com>
	<CABm2gDqmOXGZ_qxyNjFYg8qTRu4Tmo7V+zYZCt0j5RekUNVkWw@mail.gmail.com>
From: Gregory Maxwell <greg@xiph.org>
Date: Sat, 30 Sep 2017 08:54:57 +0000
Message-ID: <CAAS2fgQR48pLBPh7=yntE8iXcQSeQU22nptze76DjAUJ9A1KSQ@mail.gmail.com>
To: =?UTF-8?B?Sm9yZ2UgVGltw7Nu?= <jtimon@jtimon.cc>, 
	Bitcoin Protocol Discussion <bitcoin-dev@lists.linuxfoundation.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Cc: Daniele Pinna <daniele.pinna@gmail.com>
Subject: Re: [bitcoin-dev] Rebatable fees & incentive-safe fee markets
Precedence: list

On Sat, Sep 30, 2017 at 3:55 AM, Jorge Tim=C3=B3n via bitcoin-dev
<bitcoin-dev@lists.linuxfoundation.org> wrote:
> Gmaxwell I think what's new is that in this case, with a single tx you wo=
uld
> take out all txs with fee below 1 btc. With current rules, you would only
> remove enoguh txs for that one to fit, not empty the whole block and mine
> only a block with that single tx.

I think this is not relevant: By paying the same amount you can delay
the same transactions today.

The difference is that your 'attack' wastes less capacity-- it can be
a simple 150 weight txn rather than a collection that add up to almost
4 million weight; but it costs exactly the same.  To the extent that
this difference matters at all, I think it's an improvement.

The only argument that I see for it not being one is that it's easier
to do accidentally.  But part of the purpose of this alternative
market is to achieve an equilibrium other than the ultrabloating one;
so yes, you're going to find outcomes where the blocks are not
maximally full.

I wonder how the economics would respond if there is a PI controller
on the maximum size, so that 'lost space' in single blocks with bogon
fee transactions could be recovered if doing so didn't change the
medium timescale total. I think the paper's analysis assumes there is
no limit, but that is impractical for technical reasons (e.g. making
it impossible to budget communications and storage capacity for
nodes...).