DBM is roughly equivalent to querying in realtime (in that we wouldn't need an a-priori listing of all users), but harder (because we would need a message passing system to keep it up to date). We can just query a fast webservice, I suspect, with a small cache. We've done this before.
@noodles Loading all the PPA tokens into memory isn't an issue, because a single run only ever updates a few minutes worth of PPAs. That said, having N queries per archive is fine, so your branch isn't a problem (even though it increases query count).
The current driver for exceptional times is the late evaluation of person, which I have fixed in the attached branch. We should see how that runs and then investigate further if needed.
DBM is roughly equivalent to querying in realtime (in that we wouldn't need an a-priori listing of all users), but harder (because we would need a message passing system to keep it up to date). We can just query a fast webservice, I suspect, with a small cache. We've done this before.
@noodles Loading all the PPA tokens into memory isn't an issue, because a single run only ever updates a few minutes worth of PPAs. That said, having N queries per archive is fine, so your branch isn't a problem (even though it increases query count).
The current driver for exceptional times is the late evaluation of person, which I have fixed in the attached branch. We should see how that runs and then investigate further if needed.