Thomas Pihl

I think this is a bit dangerous. The semantics where a publish to stream can validate that no new messages was persisted in between, may already create issues in busy...

If the order is only important per subject, sender (and batch), why not send all that data in same message? Sort and group by subject and then send a msg...

Thats not what I meant. don’t filter on their side. But if order only is relevant per subject, publish one msg per subject with all msgs for that subject in...

Maybe if you give concrete example?

What I tried to explain before is that your case should be one event (the transfer) with a from, to and amount One optimization I’m assuming you do is to...

Just a quesrion; if it’s the same work needed to be done to support your use case, why do you think latency will be smaller if it’s solved inside nats...

You mean the latency induced by 2 steps vs latency induced by locking, committing, rollbacks and similar? Ofc you’re right. If we only consider one msg with two destination subjects,...

So then you don’t need to batch send, just send them async. The rollback/lock was for the ‘all or nothing”, and if that’s not needed, no issue. Or do a...

I absolutely agree that transactions, as well as exactly once, could simplify some scenarios, but at what cost? Because the more complexity you add, the more corner-cases will appear. Moving...

Thats not really what i meant. The MaxWait i am looking for is the push-consumers timeout before it will send an un-acked message. Defaults to 30 seconds, visible when you...