-
Notifications
You must be signed in to change notification settings - Fork 535
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add the ability to process locally available ops prior to returning a container #4793
Comments
Applying ops from snapshot right away is rather easy. But it will not help much the reason we have this issue - caching of snapshots and client not having data on how far it is behind without making network call. Office_Fluid_FluidRuntime_Performance 5.6 16.3 52.5 The way I see it, the only benefit of this work is in conjunction with Issue #4769. |
I certainly think it's most beneficial with #4769. One case I personally have observed this being beneficial outside of snapshot caching is for create new scenarios. I've often noticed that the first summary is not generated for quite some time, so when doing things like sharing the file or pasting the component URL into chat, you do see the "flash of ops" that is quite jarring. Granted, this could also be solved making sure the number of ops here is minimal by updating snapshotting heuristics. |
@vladsud - I've actually gotten some feedback from folks on our team that the op playback problem is not limited to snapshot caching. In particular, we are getting reports of users seeing the "blank template with placeholders" for the first render and then some op playback to get the latest changes. This is for docs the user has never opened, so snapshot caching is not in the picture. I'm wondering if in these scenarios the trees/latest call has a good chunk of relevant ops that we should process before handing back the container to the user. |
Merging into here #4793: Need to figure out how to apply ops that are coming in snapshot, and do so in generic way not to force other drivers to follow ODSP model. Some key bugs describing why it's pretty broken for users: |
There arise some scenarios where we are loading a container that puts us in a position that we have a set of ops available locally (i.e we don't need an additional network request to get these ops). Currently, this happens in the ODSP driver, which returns the latest ops on top of the latest snapshot. In the future, we may also cache these in the driver persistent storage (see #4767)
The text was updated successfully, but these errors were encountered: