Sherpa v1.0.1 Update
admin July 29, 200905:48 pmNo CommentsYes, we only launched it this morning, and there is already an update this afternoon.
Yeah, I know, I know. “What’s going on out there in Santa Monica?”
Well, it turns out that as soon as we started getting massive downloads our server logs showed a problem with one of the requests the phone was making. While it was a minor issue that only affect a very small fraction of phones–and while the client recovered immediately after making the bad request (so it didn’t affect anyone’s actual experience), it turns out the bad request was, well, bad.
And with the unanticipated popularity of this app, a minor problem creating a bad request that only affects a very small fraction of phones out there scaled up to become a major pain in the neck for our server.
So, at the last minute we decided to slip out a fix which resolved this bug.
(For the technically inclined, the fundamental problem is that during startup on a phone that has never used GPS before, there is a race condition in the state machine which manages the list of known POIs and the code which obtains the position. This race condition created a very small window where a search request gets sent to the server that doesn’t actually contain a location to look for. The request then returns “no cigar”–which causes the client to try again, but with the location.
Which isn’t a problem, except the server really didn’t like it. Of course we have more than enough capacity for our projected usage–but why flood the back-end with bad requests?)
Yes, we really tested this thing. Our crack team of testers and developers are some of the best people I have ever worked with in my 21 years as a professional software developer. But this was one of those really weird cases that only affected a very small fraction of phones that we would never have caught if we weren’t just so darn popular!





