Proposal for lazy-loading room members to improve initial sync speed and client RAM usage

[benparsons]

Documentation: https://docs.google.com/document/d/11yn-mAkYll10RJpN0mkYEVqraTbU3U4eQx9MNrzqX1U/edit
Author: @ara4n
Date: 05/03/2018
PR: #1287

[ara4n]

Implementation over at matrix-org/synapse#2970

[ara4n]

It's worth noting that lazyloading members is entirely orthogonal to paginating the room list - in an ideal world we could end up supporting both.

[ara4n]

rolling this back to proposal-review as even though there’s been informal thumbs up from @richvdh and @erikjohnston, the whole thing is blocked on #688 before it can progress to completion and the proposal should be formally reviewed in the context of #688

[richvdh]

given that I've been busily merging PRs which implement this, and I gather that client-side support is likewise basically done, it seems like this is basically being treated as ready.

In other words: I'd suggest that anyone with objections makes them quickly, and it would be nice if the doc could be updated to clarify the current state.

[richvdh]

Implementation over at matrix-org/synapse#2970

(and matrix-org/synapse#3589, matrix-org/synapse#3568, matrix-org/synapse#3567 and matrix-org/synapse#3331)