I don't know all the inner workings of Search, but I'd assume that each request is load balanced to a random set of horizontally partitioned databases. Each set of databases will tend to be at a slightly different point in replication. I would imagine that providing perfectly consistent result sets across queries would be a non-goal of such a search system. Non-real-time search engines can mask or eliminate this issue, but masking this in a continuously updated system would be a challenge. Deduplication and tolerance of gaps between pages may be part of programming against a continuously updated store.
-John Kalucki Services, Twitter Inc. On Jul 19, 3:01 am, Zac Witte <zacwi...@gmail.com> wrote: > For most of this week I have been seeing duplicate tweets appear when > I quickly paginate through a set of results using the json search api. > This only happens when making requests in quick succession. I have > verified it in my own java application trying two different json > parsers as well as this python script that someone else wrote to > detect this very issue, which has happened previously. > > http://code.google.com/p/twitter-api/issues/detail?id=655 > > this is where the issue was previously recorded and > close:http://code.google.com/p/twitter-api/issues/detail?id=646 > > Can anyone else confirm that this is not programmer error on my part? > Twitter: what's the status?