Hi there,
This is a known issue [1] that I have been working on for some time. I have a fix ready to be deployed next week so hopefully this will start working correctly. Every API user wants and deserves the correct data and the cache should work correctly to keep load off the databases and still return you the correct results. The fix for this bug is a long time coming because we wanted to be sure it would fix the issue, perform well, and be resistant to breaking again in the future. Sorry for the long delay.
Thanks; — Matt Sanford [1] - http://code.google.com/p/twitter-api/issues/detail?id=157 On Mar 26, 2009, at 05:15 PM, krwhite wrote:
I can report that it's definitely broken in XML. The XML output I see shows a random following result, which suggests to me that the entire result set is a cache from something else. If I may suggest a fix to twitter, how about an extra parameter for 'show_the_right_results' as a boolean? The command would only be slow and uncached for those who would rather have the correct following information. In the future the parameter can just be ignored when it's faster with proper caching. This command is only sent when the user wants extended user information, so it's probably issued the least out of all of them I have implemented. I would assume that's a similar case for other clients. It would be pretty rough on the server as well to use the brute force method of getting ALL followers before displaying follow/unfollow, so I don't see many options right now aside from having multiple, confusing buttons.
