Florent's 3 Nov 2011 email:

"The spec has:

<datetime string> ::= YYYY-MM-DDThh:mm:ss.sss[Z | +hh:mm | -hh:mm]

Strictly reading this BNF, this means that the timezone is optional

In my mind I'd always read that as a strict choice between the 3
possibilities, but that's not what it actually says. (And I'm not sure
but I may have been the one that wrote that BNF in the first place
thinking it meant something else, in which case I have to live with
that mistake.)

Just a FYI really, as OpenCMIS also made the wrong interpretation."