Change proposal for adding backfill, lobby based matchmaking support to Open Match

[sawagh]

Open Match's recommended matchmaking model (Match Functions generating complete matches) does not naturally lend itself well to backfill, lobby based matchmaking scenarios. Issue #1148 covers providing guidelines to implementing this scenario using existing constructs provided by Open Match.

However, based on TSC discussions, looks like the existing concepts may fall short of providing an efficient mechanism for implementing backfill, lobby based matchmaking at scale.

Please use this proposal to address the following:

• Provide details of why the current recommendation for implementing these scenarios using Open Match is not sufficient. Additional context of scale may be useful (is this OK for some users but only has issues beyond a certain scale? or for certain scenarios?)
• Propose changes to Open Match that can help support these scenarios natively as opposed to constructing them using existing concepts.

[sawagh]

Given we are a week away from v0.10 and we dont yet have a proposal, the implementation here wont make it to v0.10 - hence pushing this out to v1.0

Per the discussions around v0.10 planning, we prefer (as much as possible) to not break the v0.10 API going forward - so we should strongly consider making the proto / API changes here either additive - or backward compatible. The client written on v0.10 that does not need this enhancement should preferably not need to change.

Also, for v1.0, we should discuss in TSC, a process for marking new unstable APIs / changes. That way, even if we introduce this in v1.0, we should be able to communicate that this is a relatively new change and may not be as stable as the other API surface (giving us room to break it even post v1.0)

[calebatwd]

Update: We're still in discovery on this. We'll try and get that into the open :)

[Laremere]

Moving backfill discussions to a new issue:
#1240