Thanks for your insight! More comments are below.
On Tue, Jan 15, 2013 at 3:18 PM, Jay Kreps <[EMAIL PROTECTED]> wrote:
Yep, we're facing the same problem with respect to late data. I'd like to
see alternative solutions to this problem, but I am afraid it's an
undecidable problem in general.
Likewise the model itself doesn't necessarily work--as you get to thousands
Right. So, you're essentially trading late data for potentially lost data?
Option 2 sounds promising.