§ 02.05Core Concepts

Limitations

Quicx is a fast in-memory broker, not a durable one. No persistence, no redelivery, no retries, one global FIFO, one core. Part of the latency advantage is work Quicx is not doing — here is exactly which work.

What Quicx does not do

Quicx is a fast in-memory broker, not a durable one. Today it does not do:

Persistence
A restart loses everything queued.
RedeliveryIn progress
If a worker dies mid-task, the task is lost and the producer is not notified.
Retries
No automatic retry, no dead-lettering.
Priorities, delays, named queues
One global FIFO.
Client libraries beyond Java
dev.quicx:quicx-client is the only one. Every other language talks the wire protocol directly — it is documented and small.
Payloads above 1 KB
PROTOCOL_MAX_PAYLOAD is 1024 bytes, matching the largest PMAD size class. Oversized frames get a clean MSG_ERROR.
Multiple coresIn progress
Single-threaded today. Scale out by running several daemons behind a TCP load balancer.

HEADS UPSome of the speed is work that isn't happening

A queue that never writes to disk, never tracks an in-flight lease and never retries has less to do per task than one that does. That is a real part of why the numbers on Benchmarks look the way they do. If you need at-least-once delivery, use beanstalkd or Redis Streams.

Feature for feature

	Quicx	beanstalkd	Redis Streams	RabbitMQ
memory bound	exact, configured	none	cap + eviction	none
binary size	54 KB	~	~1 MB+	large
dependencies	none	libevent	none	Erlang/OTP
delivery	at-most-once	at-least-once	at-least-once	at-least-once
survives worker crash	no	yes (TTR)	yes	yes
persistence	no	optional	yes	yes
multi-core	no	no	partial	yes

When Quicx is the right call

Quicx is the task queue you reach for when you need to know, exactly, how much memory the broker will use. A fixed pool declared at startup, exact per-size-class accounting, and no growth under load — that is the guarantee none of the systems in the table above make.

Speed is the supporting evidence, not the headline. If your tasks are cheap, local, replayable from the source of truth, and you would rather have a hard memory ceiling than durable delivery, Quicx fits. If losing a task on a worker crash is unacceptable, it does not — and no benchmark number changes that.