Closed beaugunderson closed 12 years ago
It's not required to break them up, but it's definitely become the preferred pattern of choice mostly so that upon first connect any user experience that requires data has some to start working from immediately and can show progress. It's also more fault tolerant, instead of the whole batch failing because of one error, it can pick up with some of the work done.
Thanks, this looks great!
Also includes a change to registry.js to allow POSTs to /auth/:id/auth (so that passwords can be sent as POST data and not via the querystring).
I have an implementation question about this: is it better to break up long-running synclets (longer than 30 seconds) to a pattern that relies on external pagination and multiple runs of the synclet to retrieve all of the data?
Right now I'm handling pagination inside of the synclet and it takes a while to run on my dataset (around a minute for ~134 HTTP gets, though I plan to make it smarter for subsequent runs).
Zeo API keys available here.