This sounds great. Right now I'm just using Amazon SQS, which lacks the "blocking fetches" like Starling did. It's probably not feasible to have something similar in SQS due to the distributed nature, but it would be nice, especially for low traffic queues. The vast majority of our SQS requests don't return a message.
They use 0.25 seconds as an example timeout, but I wonder if there's any reason it couldn't be something much higher?
well, it's holding on to an active connection during that timeout, right? if it were much higher, you could quickly eat up all your available connections. on the flip side, for something low traffic that may not be an issue.
Well, I have being using starling for a private project on my own and it is quite a good piece of software. Somewhere, I think perphaps in the project page it was repported to be able to cope with >=4000 msg/sec.
They use 0.25 seconds as an example timeout, but I wonder if there's any reason it couldn't be something much higher?