After spending most of yesterday trying to understand what was going wrong with the Flickr API, I decided to look in a different direction and tested the uploading code, and it didn’t work either, so I posted a note on the yws-flickr mail list, hoping that someone else would have an idea how to proceed. A few minutes later John Watson posted a suggestion. I haven’t tried his workaround yet, but I will, probably after lunch.
Don Park is having the same problem with his Flickr app.
Finally, we (seem to have) solved the problem. Thanks to Don and John for their help. Here’s a screen shot of the routine that implements the technique we’re all using. Basically it now takes more time for the data to move around on the Flickr side. Is this breakage? Hard to say — it is communication software, so you have to be prepared for latency and packets getting dropped. It did used to work. So it’s both breakage and more robustness was required on our side. It just shows how sensitive these systems can be, and of course that they used a fancier method for security, that is safer, but also more fragile. There are always tradeoffs, that’s one of the constants of computer system design.