[rabbitmq-discuss] AMQP Field Tables & Short Integers

[Simon MacMullen]

On 17/01/2014 03:11, Tony Garnock-Jones wrote:
> On 16 January 2014 19:57, Gavin M. Roy <gavinmroy at gmail.com
> <mailto:gavinmroy at gmail.com>> wrote:
>     Will RabbitMQ be staying with the misuse of /s/ for shorts for the
>     foreseeable future?
>
>
> I can't speak for the team, of course, but I'd recommend doing so. All
> the client libraries I'm aware of ignore 0-9-1 in this area.

Yes. I'm not sure I'd say misuse; it's almost like a typo in the spec 
(see also: "AMQP1191" to negotiate *0-8*). And changing this would cause 
everybody's clients to break, or need to work differently depending on 
the declared server version. Can open; worms everywhere.

> If an AMQP 0-9-2 is ever released, IMO it should reflect what the
> various libraries actually implement.

Very much so!

>     Is there any intention or desire to add the specified /U/ into the
>     decoder and never support /s/ for anything but short-int instead of
>     short-string?
>
>
> This sounds like a good idea.

I'm not so sure. It doesn't really bring compliance with the letter of 
the spec, it doesn't let you do anything new, but it does muddy the 
waters even further. At least at the moment we can say "these are the 
types that are in practice used everywhere".

Cheers, Simon