Message-Id: <1491681378.2454247.938587616.7199D633@webmail.messagingengine.com>
From: Tomas <tomas@tomasvdw.nl>
To: Johnson Lau <jl2012@xbt.hk>,
	"bitcoin-dev" <bitcoin-dev@lists.linuxfoundation.org>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"
In-Reply-To: <6F1E6FB6-1342-4BD6-BF83-A160C1A7CD34@xbt.hk>
Date: Sat, 08 Apr 2017 21:56:18 +0200
References: <1491516747.3791700.936828232.69F82904@webmail.messagingengine.com>
	<CAAS2fgTJ8xOj8zCmBq1LN9OdMV-tDfSjVUPhLpO98cR1w-QAoA@mail.gmail.com>
	<CA+KqGko0cDY29bhznMxJJ7yAUTuB6GaDDNGBRwzssJUxM_53xQ@mail.gmail.com>
	<1491599691.1245876.937920664.6EBA20DC@webmail.messagingengine.com>
	<CAAS2fgTWyX5M-xcELC2vDvGfs01tbGYkpZJCSeNbvn_p4Ecjqg@mail.gmail.com>
	<1491636528.2474173.938219072.54C44183@webmail.messagingengine.com>
	<6F1E6FB6-1342-4BD6-BF83-A160C1A7CD34@xbt.hk>
Subject: Re: [bitcoin-dev] Using a storage engine without UTXO-index
Precedence: list


> I don=E2=80=99t fully understand your storage engine. So the following de=
duction
> is just based on common sense.
>=20
> a) It is possible to make unlimited number of 1-in-100-out txs
>=20
> b) The maximum number of 100-in-1-out txs is limited by the number of
> previous 1-in-100-out txs
>=20
> c) Since bitcrust performs not good with 100-in-1-out txs, for anti-DoS
> purpose you should limit the number of previous 1-in-100-out txs.=20
>=20
> d) Limit 1-in-100-out txs =3D=3D Limit UTXO growth
>=20
> I=E2=80=99m not surprised that you find an model more efficient than Core=
. But I
> don=E2=80=99t believe one could find a model that doesn=E2=80=99t become =
more efficient
> with UTXO growth limitation.

My efficiency claims are *only* with regards to order validation. If we
assume all transactions are already pre-synced and verified, bitcrust's
order validation is very fast, and (only slightly) negatively effected
by input-counts.

Most total time is spend during base load script validation, and UTXO
growth is the definitely the limiting factor there, as the model here
isn't all that different from Core's.


> Maybe you could try an experiment with regtest? Make a lot 1-in-100-out
> txs with many blocks, then spend all the UTXOs with 100-in-1-out txs.
> Compare the performance of bitcrust with core. Then repeat with
> 1-in-1-out chained txs (so the UTXO set is always almost empty)
>=20

Again, this really depends on whether we focus on full block validation,
in which case the 100-1, 1-100 distinction will be the similar to Core,
or only regard order validation, in which case Bitcrust will have this
odd reversal.=20


> One more question: what is the absolute minimum disk and memory usage in
> bitcrust, compared with the pruning mode in Core?

As bitcrust doesn't support this yet, I cannot give accurate numbers,
but I've provided some numbers estimates earlier in the thread.


Rereading my post and these comments, I may have stepped on some toes
with regards to SegWit's model. I like SegWit (though I may have a
slight preference for BIP140), and I understand the reasons for the
"discount", so this was not my intention. I just think that the reversal
of costs during peak load order validation is a rather interesting
feature of using spend-tree  based validation.=20

Tomas