[jdev] XEP-0115: backward compatibility

[Remko Tronçon]

>  1. If I receives a presence that contains both a 'hash' and 'ext'
>  attributes like this:

If your client supports hash, and hash is present, then you have to
ignore ext. I don't think it makes any sense for a client to send
'ext' and 'hash' at the same time (see below).

>  But a client compliant to the XEP version 1.3 (where is defined the
>  legacy format) will send a discovery query on the ver attribute,
>  whatever it contains (hash as specified in v1.5 or version as in v1.3).
>  It seems inconsistent.

That section is about the *processing* side: if you *receive* a 'ver'
without a hash, then you should not try to discover it (since you
don't support it, and you can't verify it), unless you implemented the
legacy algorithm as well. Either way, you will *always* support
replying to ver queries, as both legacy and new protocol use this.

>  3. If a client with pluggable capabilities want to be compatible with
>  both remote client implementing the XEP v1.3 and v1.5, is the
>  recommended way to use both 'ext' and 'hash' attributes?

I don't think combining both makes sense (or will even work properly).
If you support the new protocol, use the new one only, and do not send
any 'ext'.

cheers,
Remko