Since discussion around allowing unknown messages or not allowing them seems contentious, I'd like to offer up another possibility: create a single new message, XVERSION, (and bump the protocol rev) which is a key-value array of arbitrary data. Any protocol extension can then choose a new key (with a 32 or 64 bit keyspace you can basically hand out prefixes to any implementation that wants one) and publish custom data via this message without needing to bump the protocol rev field. Typical "custom data" would be the min and max supported version of some specific extended protocol, but any data is possible since the "value" field can be serialized via the same network serialization format. It therefore doubles as a "configuration" message as well as protocol extension negotiation. For example, we use it to communicate the maximum unconfirmed chain a node will commit to the mempool, and peers don't bother to send transactions that exceed this limit. You can find a specification here: https://gitlab.com/bitcoinunlimited/BCHUnlimited/-/blob/dev/doc/xversionmessage.md Code has been deployed for a long time. Regards, Andrew On Mon, Aug 24, 2020 at 5:44 AM Suhas Daftuar via bitcoin-dev < bitcoin-dev@lists.linuxfoundation.org> wrote: > Hi all, > > Thanks for the helpful discussion. > > My primary motivation in starting this thread was to establish what the > expectations are for new feature deployment (particularly whether the > protocol version should continue to be bumped or not), and I think I have > that answer -- different from what I proposed when I started this thread, > but not in a way that I think meaningfully hinders future work. So I'm > happy to leave it at that and withdraw my suggestion. > > Cheers, > Suhas > > > On Sun, Aug 23, 2020 at 1:51 PM Eric Voskuil via bitcoin-dev < > bitcoin-dev@lists.linuxfoundation.org> wrote: > >> >> > On Aug 21, 2020, at 15:16, Matt Corallo >> wrote: >> > >> > Hmm, could that not be accomplished by simply building this into new >> messages? eg, send "betterprotocol", if you see a verack and no >> "betterprotocol" from your peer, send "worseprotocol" before you send a >> "verack". >> > >> > Matt >> > >> >> On 8/21/20 5:17 PM, Jeremy wrote: >> >> As for an example of where you'd want multi-round, you could imagine a >> scenario where you have a feature A which gets bugfixed by the introduction >> of feature B, and you don't want to expose that you support A unless you >> first negotiate B. Or if you can negotiate B you should never expose A, but >> for old nodes you'll still do it if B is unknown to them. >> >> This seems to imply a security benefit (I can’t discern any other >> rationale for this complexity). It should be clear that this is no more >> than trivially weak obfuscation and not worth complicating the protocol to >> achieve. >> >> >> An example of this would be (were it not already out without a feature >> negotiation existing) WTXID/TXID relay. >> >> The SYNC primitve simply codifies what order messages should be in and >> when you're done for a phase of negotiation offering something. It can be >> done without, but then you have to be more careful to broadcast in the >> correct order and it's not clear when/if you should wait for more time >> before responding. >> >> On Fri, Aug 21, 2020 at 2:08 PM Jeremy > jlrubin@mit.edu>> wrote: >> >> Actually we already have service bits (which are sadly limited) >> which allow negotiation of non bilateral feature >> >> support, so this would supercede that. >> >> -- >> >> @JeremyRubin < >> https://twitter.com/JeremyRubin> >> _______________________________________________ >> bitcoin-dev mailing list >> bitcoin-dev@lists.linuxfoundation.org >> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev >> > _______________________________________________ > bitcoin-dev mailing list > bitcoin-dev@lists.linuxfoundation.org > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev >