Received-SPF: pass (sog-mx-2.v43.ch3.sourceforge.com: domain of bitpay.com
	designates 74.125.82.177 as permitted sender)
	client-ip=74.125.82.177; envelope-from=jgarzik@bitpay.com;
	helo=mail-we0-f177.google.com; 
MIME-Version: 1.0
In-Reply-To: <CABsx9T2BDBNqvinVNk3FmBRWU7R8jf6Vm6NaH74te0FRCh1O-w@mail.gmail.com>
References: <CA+iPb=EaX=bvOjNtZ+LnYTMRLQQ9nFcrefAkBdv8eActoX_b8A@mail.gmail.com>
	<CABsx9T0ag_o_mu=5Q7Ju7s2hO3jz-o5g9FihE9h4B6+ednd2Pg@mail.gmail.com>
	<CAJHLa0NZRF+1QjSwtwjaTE07NWJ_U-O-DE24=P5eSAutMqTupg@mail.gmail.com>
	<CABsx9T2BDBNqvinVNk3FmBRWU7R8jf6Vm6NaH74te0FRCh1O-w@mail.gmail.com>
From: Jeff Garzik <jgarzik@bitpay.com>
Date: Fri, 18 Jul 2014 11:06:23 -0400
Message-ID: <CAJHLa0O=eCoyvV19dWgTnYd9Di0wLLZtWmCPidc-dWqPNQv_oQ@mail.gmail.com>
To: Gavin Andresen <gavinandresen@gmail.com>
Content-Type: text/plain; charset=UTF-8
Cc: Bitcoin Dev <bitcoin-development@lists.sourceforge.net>
Subject: Re: [Bitcoin-development] Squashing redundant tx data in blocks on
 the wire
Precedence: list

On Fri, Jul 18, 2014 at 10:53 AM, Gavin Andresen
<gavinandresen@gmail.com> wrote:
> But if there was some agreed-upon canonical ordering, then it should
> theoretically be possible to take shortcuts in the "what order".
>
> You'd start with setof(transactions I think everybody knows about)
> Select some subset, based on miner's policy
> Sort that subset with the canonical ordering algorithm
> Very efficiently broadcast, taking all sorts of shortcuts assuming most of
> your peers already know the set you started with and expect the same
> canonical ordering (see gmaxwell's thoughts on block encoding).

Related implementation detail:  Having pursued this train of thought,
I noted that you don't want to include too-young transactions that you
received in the past few seconds, because those are likely still
propagating around the network.

> Second half-baked thought:
> I wonder if broadcasting your transaction selection policy ("11KB of free
> transactions, sorted by priority, then 111K of fee-paying transactions,
> sorted by fee") might make it possible to save even more bandwidth by
> letting your peers create a very good approximation of your block with just
> that information....

Absolutely.  One path I would like to see pursued is multiple
p2pool-esque chains.  Each with their own policy, perhaps with their
own administrative team.  ie. you could have a fully decentralized
p2pool-like chain, or multiple such chains, each with a stated
policy/reward pattern.  Or, GHash/BTCGuild/Eligius could run a
semi-centrally managed chain ultimately guaranteed not only by
protocol but by administrators' digital signatures.

In each case, advertising technical attributes about your pool [chain]
policy would give nodes the better ability to predict what is in an
upcoming block.

And the flip side of that, such predictions are never perfect.  Need
to make sure the fallback case, while undoubtedly more costly than the
Fast Path, is not overly painful.


-- 
Jeff Garzik
Bitcoin core developer and open source evangelist
BitPay, Inc.      https://bitpay.com/