References: <CA+8xBpe9D=poPyJ=soGdN3sovqdmvyGGij6FM8PHYGUB5aUkzQ@mail.gmail.com>
Message-ID: <1364035346.19716.YahooMailNeo@web161601.mail.bf1.yahoo.com>
Date: Sat, 23 Mar 2013 03:42:26 -0700 (PDT)
From: Randy Willis <willis.randy@ymail.com>
To: Bitcoin Development <bitcoin-development@lists.sourceforge.net>
In-Reply-To: <CA+8xBpe9D=poPyJ=soGdN3sovqdmvyGGij6FM8PHYGUB5aUkzQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
Subject: Re: [Bitcoin-development] A bitcoin UDP P2P protocol extension
Precedence: list
Reply-To: Randy Willis <willis.randy@ymail.com>

Not looked at code yet, some thoughts:=0A=0A1) In IPv4 max UDP data size is=
 65507 bytes.=0A2) What about big messages (block)?=0A3) I think relay spee=
d can be increased only by reducing network diameter. =0A=0AIntroducing sup=
er-nodes with thousands of connected peers can greatly help here.=0A4) The =
only (IMO) UDP advantage: hole punching. But requirement of active =0A=0ATC=
P connection defies it. And hole punching will need protocol extension.=0A5=
) unsolicited tx broadcast will increase traffic of nodes (especially well-=
connected)=0Aand less people will run them.=0A=0A=0A---- Original Message -=
----=0A=0AFrom: Jeff Garzik <jgarzik@exmulti.com>=0ATo: Bitcoin Development=
 <bitcoin-development@lists.sourceforge.net>=0ACc: =0ASent: Saturday, March=
 23, 2013 7:17 AM=0ASubject: [Bitcoin-development] A bitcoin UDP P2P protoc=
ol extension=0A=0AHere is a rough draft implementation of a UDP P2P protoco=
l extension=0Afor bitcoin:=0A=0A=A0 =A0 https://github.com/jgarzik/bitcoin/=
tree/udp=0A=A0 =A0  http://yyz.us/bitcoin/udp-v0.patch=0A=0AProtocol specif=
ication (such that it is):=0A=0A- UDP, bound to same port as TCP P2P (norma=
lly 8333)=0A- Active, simultaneous TCP P2P connection required (useful agai=
nst DoS=0Aand other attacks)=0A- Same message format as TCP P2P, same pchMe=
ssageStart conventions,=0Aetc. (my CNetMessage pull req would be helpful he=
re)=0A- Multiple P2P messages per UDP packet permitted=0A- Max UDP packet s=
ize 100*1024 bytes=0A- Advertises NODE_UDP in nServices=0A- New "getudpcook=
" TCP P2P command returns a "udpcook" message,=0Acontaining a yummy cookie=
=0A- UDP P2P commands "inv", "tx" and "addr" are handled as if received via=
 TCP=0A- UDP P2P command "udpsub" sets a mask, that subscribes to one or mo=
re=0Adata broadcasts.=0A- When USM_INV_BCAST mask bit is set, receive "inv"=
 messages via UDP=0Arather than TCP.=0A=0AProject and design goals (or, why=
 do this?):=0A=0A- It is theorized that UDP may be useful for some messages=
 we=0Abroadcast throughout the network=0A- A productive discussion STARTS w=
ith code, otherwise discussion=0Acontinues forever.=0A- Investigate improvi=
ng "inv" relay speed=0A- Investigate unconditional "tx" broadcasting via UD=
P, as "inv"=0Aalternative, for small tx's.=0A- Investigate improving block =
relay speed (or perhaps block header relay speed)=0A- Open up new design av=
enues, for P2P patterns more suited to UDP than=0ATCP in general.=0A=0A-- =
=0AJeff Garzik=0AexMULTI, Inc.=0Ajgarzik@exmulti.com=0A=0A-----------------=
-------------------------------------------------------------=0AEveryone ha=
tes slow websites. So do we.=0AMake your web apps faster with AppDynamics=
=0ADownload AppDynamics Lite for free today:=0Ahttp://p.sf.net/sfu/appdyn_d=
2d_mar=0A_______________________________________________=0ABitcoin-developm=
ent mailing list=0ABitcoin-development@lists.sourceforge.net=0Ahttps://list=
s.sourceforge.net/lists/listinfo/bitcoin-development=0A