[zeromq-dev] Wire-format spec for zmq 3.2

KIU Shueng Chuan nixchuan at gmail.com
Wed Oct 24 10:00:22 CEST 2012

On Wed, Oct 24, 2012 at 2:58 PM, Martin Hurton <hurtonm at gmail.com> wrote:
>> I think (would have to check the code) that 2.x peers do not consider
>> the MORE bit on identities.
> Yes, the 2.x never checks that bit.

Since all flag bits except bit 0 are reserved, an ZMTP/1.0
implementation is supposed to ignore those bits. So sending %7E
instead of %7F for the handshake would not rely on the implementation
ignoring the MORE bit.

> Encoding the frame length in 9 bytes for short identities (0-254
> bytes) is another violation of ZMTP/1.0 spec.
> Again, zeromq 2.x implementation does not care, but may be problem
> with more strict implementations.

Would it work if the language of the ZMTP/1.0 spec could be relaxed
from SHALL to SHOULD concerning encoding of lengths?
i.e. An implementation should only send short lengths using 1-byte
encoding but if it encounters 9-byte encodings, it should accept them.

More information about the zeromq-dev mailing list