Maximum AllWatcher frame size?

[roger peppe]

On 16 June 2017 at 06:03, John A Meinel <john.meinel at canonical.com> wrote:
> Wouldn't that feel like a breach of contract where the first response is
> supposed to get you in sync with the system?

Yes, that occurred to me too, and it is a potential issue.
We could provide some indicator that it was incomplete, perhaps,
or the other way around - an indicator that it *is* complete.

> I could definitely see doing a different code path for the first response
> because you know there aren't any objects you have to compare against, etc.
> Or because you can reasonably expect that the first response is going to be
> an entirely different order of magnitude to everything else. And at some
> point you do have to load in the whole model to get started.
> Especially bad we ended up with something like the 5s watcher delay between
> updates and the initial model load of a large model takes 100s because we
> broke it up.

The main overhead to sending several batches is the extra round trips,
but I suspect that wouldn't be too bad in practice - we should only
be seeing extra round trips when the overall size is very large anyway.
Note that the usual 5s watcher delay between updates is not inherent - it's
only incurred for synchronisation events between mongo and the
in-memory store.

  cheers,
    rog.