[Cbor] correctness of implied top level array?

[Laurence Lundblade <[e-mail]>]

The current CBOR playground works like this:

    Input 0x00, (one byte) and it decodes to a single integer of value 0
    Input 0x82 0x00 0x00 (three bytes) and it decodes to an array of two zeros
    Input 0x00 0x00 (two bytes) and it decodes to a single integer with one extra byte, an error
    Similarly substitute 0x40, a zero length byte string, for 0x00 and it works the same as the above three
    Similarly substitute 0xf4, “false", for 0x00 and it works the same above three

That is, the top level of a chunk of encoded CBOR can either be a map, an array or a single non-structured data item like an integer, string or simple type. 

I think in theory you could have an implied array at the top level and allow 0x00 0x00 to decode correctly, but it is probably better not to allow this implied array. It might save someone a byte or two or three or five or nine, but I think allowing it creates an unnecessary special case. 

So 0x00, 0x00 and such should be documented as not-well-formed in the RFC and the Appendix C pseudo code should be fixed by tracking the whether the top-level is being decoded or not.

LL