On 12/01/2015 14:12, Matthew White wrote:
> Hello;
>
> Just occurred to me that we've been discussing this for BytesMessages - does the same not apply to Stream messages as well?
>
> They have the same read/write semantics - and if one is being clarified should the other?
>
Yes, very good point! I think we should indeed make exactly the same change to StreamMessage. Exactly the same arguments
apply.
I checked the CTS tests for async send with a StreamMessage and confirmed that these are similar to the tests for
BytesMessage. They read (using readBoolean) the body of the message passed to the CompletionListener without calling
reset() beforehand. This means that, although the CTS tests are not part of the spec, we can be sure that existing
implementations already call reset() on the StreamMessage passed to the CompletionListener.
Comments welcome. Meanwhile I'll formally draft the required changes.
Nigel