At first glance, it seems that you should track on the keywords you care about with the Streaming API, and then sort by user on your client end.
The count parameter allows clients to mask data loss at reconnection- time. It won't be sufficient for your purposes as the look back is only a few tens of minutes at most. -John Kalucki http://twitter.com/jkalucki Services, Twitter Inc. On Sep 8, 10:17 pm, LeeS <lse...@gmail.com> wrote: > For a new project we'll need to retrieve the text of recent statuses > for a large group of specific users (several thousand to start), > matching a smaller list of keyword strings. Both the users and > keywords will grow over time but the keyword set will probably remain > at least an order of magnitude smaller than the set of users and grow > more slowly Real time is not important and the statuses can be a few > hours old. Is using the 'shadow' method of the Streaming API with a > negative 'count' parameter the best way to do this with a minimum load > on the API? If so, what's the best way to obtain the access needed? > It looks like the initial base of users is going to outstrip the > 'filter' method's limitation, and 'filter' doesn't allow for the use > of 'count' in the default role. > > Thanks > > --Lee Semel > > Recent Twitter projects: http://muckrack.com -http://venturemaven.com > -http://twittorati.com-http://shortyawards.com-http://shortyawards.com/twitter_pro/