Raffi Krikorian
summary
Twitter moves from per-user geodata to per-tweet geotagging and faces a scaling problem that requires a new service. A reverse geocoder called Rockdove runs RPCs to convert lat/long to place hierarchies, do IP lookups, and expose CRUD for places, all with unauthenticated calls and rate limits. Front end, API, search and Hadoop analytics all hit the same normalized store so lookups happen at tweet ingestion speed and the data is usable both internally and externally.
AI-written summary · about
related talks
39mRoasting Google APIs on a Firebasse to create Iberritos
16mA Conversation with Ryan Sarver and Bret Taylor
40mArchitectures for Exploring Your Personal Data
28mWhat is wrong with API wrappers and how can we do better
30mBowerbirds of Technology: Architecture and Teams at Less-than-Google Scale
30mData Rainbows - select * from cloud;
11mWeb 2.0 Expo NY 2010: Dennis Crowley
34mLightning Talks