Talk:Paxos (computer science)

Is the diagram for Message flow: Basic Paxos, failure of redundant Learner accurate?
The indicated return from the acceptors is: Promise(1,{null,null,null}) Should not the return from acceptors instead be: Promise(1,{Va,Vb,Vc}) ? ie they are successful.

I believe the first failure in this scenario is later on with the Learner. — Preceding unsigned comment added by Davidberkowitz9 (talk • contribs) 23:17, 13 October 2014 (UTC)

Is the section "Paxos can be used to select a leader" accurate?
I don't think there is a "winning" Proposer in Paxos. Even after consensus has been reached, another Proposer can always come along and pass another proposal through the system (with the same value, of course). Leader election could easily be done by incorporating the initial proposer's network identifier into the value, but this is not part of Paxos proper and not what is described in the section. — Preceding unsigned comment added by CobaltThoriumG (talk • contribs) 02:55, 3 December 2018 (UTC)

Replace text diagrams with images?
ASCII diagrams representing the communication process are hard to read, wouldn't it be better to replace them with graphics? 213.252.196.189 (talk) 10:08, 11 September 2008 (UTC)Vova D.

I've found an online sequence diagram generator, that with a bit of effort, I think someone could generate equivalent sequence diagrams that would be kinder on the eyes. Linearizable (talk) 16:49, 14 October 2017 (UTC)

I agree in general that this article should use images. As a student trying to understand the Paxos protocol, I find all of the papers explaining it confusing especially when I only have text to go by. Qzekrom (she/they &bull; talk) 06:34, 5 November 2019 (UTC)


 * I think we can use PlantUML to get pretty decent results in a high quality SVG format. The diagrams end up much larger, but are probably clearer anyway.  Here's the link to the text with a preview of the Basic Paxos success case.  And here is the resulting image.  Ideally we could capture both the text that generates the graph too, so that it's easy to make modifications later if needed. Iamfromspace (talk) 00:29, 28 May 2022 (UTC)
 * Just replaced the first sequence diagram with an SVG. Planning to migrate the rest a little at a time.  I'll also let this first one sit for a bit in case there's any other feedback that people want to share here. Iamfromspace (talk) 03:39, 9 June 2022 (UTC)

More explanation in the beginning
The article just jumps into things. What is a learner? —Preceding unsigned comment added by 98.229.153.38 (talk) 04:51, 10 June 2010 (UTC)

Ascii graphics issues
The ascii graphics do not make it obvious by their vertical lanes whether a proposers sends its proposal via a broadcast to all acceptors (likely), or the message gets forwarded from one acceptor to the next. Should be rendered in a more explicit way... Cheers, [ip] — Preceding unsigned comment added by 141.62.31.244 (talk) 08:19, 13 January 2012 (UTC)

Virtual Synchrony and its relationship to Paxos
Early this year there was discussion of reconfigurable Paxos and the connections between that form of stoppable state machines with other ways of building reconfigurable protocols strong enough to do state machine replication. People interested in this topic might find the Gbcast Protocol page interesting.

Gbcast (these days, sometimes known as "virtually synchronous Paxos") is a protocol from the early Isis Toolkit. The Wiki page shows how that protocol bisimulates with basic Paxos, and also outlines a step-by-step process for turning Basic Paxos into Gbcast, or vice versa. 128.84.154.13 (talk) 16:19, 12 November 2012 (UTC)

Great article
Thanks to everyone that contributed. Very helpful. Dan 12:38, 26 September 2012 (UTC)

Links partly not working
Hy, I could not figure out how to correct the links. At the "Phase 1a: Prepare" section, the Proposer and Leader link is not working. How can you add html-anchors in WikiMarkup?

There's a anchor template that does this. I've added it to the needed places to get the links to work. Thanks for pointing out the issue! Linearizable (talk) 17:05, 14 October 2017 (UTC)

Paxos Variants
I'd like to expand this article to cover additional paxos variants, their advantages, and link out to the primary research done. Ideally, I'd like to cover the same material as the overview of the variants in Paxos Made Moderately Complex [CSUR 2015]. Is just adding more sections to this article the best place to do so? Linearizable (talk) 16:57, 14 October 2017 (UTC)