MIME-Version: 1.0
In-Reply-To: <CALwsPgm9S3UNd3bEuWreyGS7bcvSD+cXxueoD+F_D9fC=xLz2Q@mail.gmail.com>
References: <CALwsPgnnkbfUBhL=Qyspz13pnZ-6RHdaZOGvfLG34JjJRgt2Dw@mail.gmail.com>
	<B3546CB9-6A24-474C-8B56-9B1E2D33B470@mattcorallo.com>
	<CALwsPgm6xcBfLXZTNTZ40R_s3oUawE0ANZycDWpSo0cXZ+=-Vg@mail.gmail.com>
	<CAAS2fgQLNH+rivNNTFz6xt_9SxO3fFj7-3z7A-_B_2X2x-6M5w@mail.gmail.com>
	<CALwsPgm9S3UNd3bEuWreyGS7bcvSD+cXxueoD+F_D9fC=xLz2Q@mail.gmail.com>
Date: Wed, 5 Aug 2015 22:14:07 +0000
Message-ID: <CAAS2fgSXWzCv=4cF=0bwL9+udzBHSPR7goL3U_c1NjS22dpWzQ@mail.gmail.com>
From: Gregory Maxwell <gmaxwell@gmail.com>
To: Arnoud Kouwenhoven - Pukaki Corp <arnoud@pukaki.bz>
Content-Type: text/plain; charset=UTF-8
Cc: Arnoud Kouwenhoven - Pukaki Corp via bitcoin-dev
	<bitcoin-dev@lists.linuxfoundation.org>
Subject: Re: [bitcoin-dev] Idea: Efficient bitcoin block propagation
Precedence: list

On Wed, Aug 5, 2015 at 9:19 PM, Arnoud Kouwenhoven - Pukaki Corp
<arnoud@pukaki.bz> wrote:
> Thanks for this (direct) feedback. It would make sense that if blocks can be
> submitted using ~5kb packets, that no further optimizations would be needed
> at this point. I will look into the relay network transmission protocol to
> understand how it works!
>
> I hear that you are saying that this network solves speed of transmission
> and thereby (technical) block size issues. Presumably it would solve speed
> of block validation too by prevalidating transactions.


Correct. Bitcoin Core has cached validation for many years now... if
not for that and other optimizations, things would be really broken
right now. :)

> Assuming this is all
> true, and I have no reason to doubt that at this point, I do not understand
> why there is any discussion at all about the (technical) impact of large
> blocks, why there are large numbers of miners building on invalid blocks
> (SPV mining, https://bitcoin.org/en/alert/2015-07-04-spv-mining), or why
> there is any discussion about the speed of block validation (cpu processing
> time to verify blocks and transactions in blocks being a limitation).

I'm also mystified by a lot of the large block discussion, much of it
is completely divorced from the technology as deployed; much less what
we-- in industry-- know to be possible. I don't blame you or anyone in
particular on this; it's a new area and we don't yet know what we need
to know to know what we need to know; or to the extent that we do it
hasn't had time to get effectively communicated.

The technical/security implications of larger blocks are related to
other things than propagation time, if you assume people are using the
available efficient relay protocol (or better).

SPV mining is a bit of a misnomer (If I coined the term, I'm sorry).
What these parties are actually doing is blinding mining on top of
other pools' stratum work. You can think of it as sub-pooling with
hopping onto whatever pool has the highest block (I'll call it VFSSP
in this post-- validation free stratum subpooling).  It's very easy to
implement, and there are other considerations.

It was initially deployed at a time when a single pool in Europe has
amassed more than half of the hashrate. This pool had propagation
problems and a very high orphan rate, it may have (perhaps
unintentionally) been performing a selfish mining attack; mining off
their stratum work was an easy fix which massively cut down the orphan
rates for anyone who did it.  This was before the relay network
protocol existed (the fact that all the hashpower was consolidating on
a single pool was a major motivation for creating it).

VFSSP also cuts through a number of practical issues miners have had:
Miners that run their own bitcoin nodes in far away colocation
(>100ms) due to local bandwidth or connectivity issues (censored
internet); relay network hubs not being anywhere near by due to
strange internet routing (e.g. japan to china going via the US for ...
reasons...); the CreateNewBlock() function being very slow and
unoptimized, etc.   There are many other things like this-- and VFSSP
avoids them causing delays even when you don't understand them or know
about them. So even when they're easily fixed the VFSSP is a more
general workaround.

Mining operations are also usually operated in a largely fire and
forget manner. There is a long history in (esp pooled) mining where
someone sets up an operation and then hardly maintains it after the
fact... so some of the use of VFSSP appears to just be inertia-- we
have better solutions now, but they they work to deploy and changing
things involves risk (which is heightened by a lack of good
monitoring-- participants learn they are too latent by observing
orphaned blocks at a cost of 25 BTC each).

One of the frustrating things about incentives in this space is that
bad outcomes are possible even when they're not necessary. E.g. if a
miner can lower their orphan rate by deploying a new protocol (or
simply fixing some faulty hardware in their infrastructure, like
Bitcoin nodes running on cheap VPSes with remote storage)  OR they can
lower their orphan rate by pointing their hashpower at a free
centeralized pool, they're likely to do the latter because it takes
less effort.