launchpad-dev team mailing list archive

[Robert Collins <robert.collins@xxxxxxxxxxxxx>]

One thing (of many :)) that I don't understand here.

You are saying that figuring out what packages need new diffs made
will be very expensive (so you can't just run in a check-and-do mode.

But you're also saying that storing a simple index into the
work-to-be-done is all the abstraction you need? That seems to imply
that generating a list of packages that need diffs made after your
timestamp/db-id is cheap to do.

These two things seem to be in contradiction to each other.

Separately, as far as modelling stuff now to fit e.g. rabbit later,
model a queue which you put work into and dispatch work from : its
what rabbit does, and what the existing TaskSystem does. (I'd probably
use the TaskSystem for anything new-and-urgent).

-Rob