Thinking about AWS SNS, which does app-to-app fanout. Only one of those is going to be a response for an interactive user. If that one drops after 5s, oh well, that person has moved on. The rest need SQS, retrying a few times and giving up seems wrong.
Another pattern I see in AWS is just how many services have a message queue hidden in them:
SNS- short queue, 50 retry attempts
Step Functions- 1 year queue!
SQS- 2 weeks
Kinesis- I still don't understand kinesis. SQS but for streaming data?♡ 2 ↻ 0Maybe a queue isn't/shouldn't be it's own thing, it should be a feature of all cloud things.
[service] -> [queue] -> [service]
vs
[service/w queue] -> [service/w queue]♡ 3 ↻ 0