tumblr gear

core concept

• text-based scraping engine. (not dom-based. and almost all rules are written in regular expressions.)
• metadata for scraping is located on outside of application. (s3)
• dynamically create a page from tumblr html with *minimum* processing.

internal architecture

• background loading queue. (nsoperationqueue. serial operation.)
• background posting queue. (nsoperationqueue. parallel operation.)
• 2-layer cache. (html cache, webview cache.)
• reusable & gradual warmup webview cache. (3 upwards, 1 current, 6 downwards. total 10 views. saving memory usage.)
• metadata for scraping. (reading & writing through nsmutabledictionary.)
• additional webview for browsing pages outside tumblr.
• notification popup. (simple uiview & quartzcore.)
• monitoring availability of network.

future plans

• basic functions will be preferentially enhanced. (stability, performance, disaster tolerance and so on. then advanced functions.)
• will be open-sourced.
• metadata will be moved from s3 to wedata.