Test 'minimal publish' has non-conforming results

[neacsum]

According to this test, the generated packet is:

  48, 10, // Header
  0, 4, // Topic length
  116, 101, 115, 116, // Topic (test)
  116, 101, 115, 116 // Payload (test)

This is completely missing the PUBLISH packet properties. However, according to MQTT V5.0 2.2.2.1 If there are no properties, this MUST be indicated by including a Property Length of zero [MQTT-2.2.2-1]

EDIT: I see that at least one server (test.mosquitto.org) accepts the message in this form but I'd like to have at least a clarification how is this standard conforming.
-- END EDIT

Shouldn't this be:

  48, 11, // Header
  0, 4, // Topic length
  116, 101, 115, 116, // Topic (test)
  0,  //PUBLISH Properties
  116, 101, 115, 116 // Payload (test)

[robertsLando]

I didn't developed that part of the code but the specs seems very clear in this, maybe most brokers are not that strict and also accept as valid packets without publish properties length. I don't see any issue BTW in accepting such packets, unfortunately I can't give you a better explanation, @seriousme how do you handle this in opifex? @mcollina any clue?

[seriousme]

@robertsLando

The Opifex encoder adds the zero for empty properties as per spec [MQTT-2.2.2.1]
The Opifex decoder will reject the packet without the zero.

See: https://github.com/seriousme/opifex/blob/mqttv5/mqttPacket/publish.ts

Kind regards,
Hans

[robertsLando]

@neacsum would you like to create a PR to add the missing packet length 0 when generating a packet? I would still keep the parser accepting packets without this to prevent breaking changes.

[seriousme]

My suggestion would be to at least log a warning when a non-compliant packet is accepted.
(or make it an option to accept non-compliant clients etc)
Postels law sounds great on first glance but over the longer term often results in a patchwork of non-interoparable implementations.

My 2cts,
Hans

[neacsum]

@robertsLando I'm afraid I cannot do it. My JS skills are very limited :) I was perusing the test cases as I was writing my own C++ MQTT client.

[seriousme]

@robertsLando
it seems that the same mechanism is used by more packets. I can do a PR if you want.

Kind regards,
Hans