Re: [core] Designs to resolve streaming issues in SenML

[Carsten Bormann <[e-mail]>]

Indeed, SenML -01 is optimized for implementations that ingest the
entire SenML document and then work on the result -- exactly what
happens in your Python implementation.

What -02 was trying to do was to also enable implementations operating
sequentially (what somewhat misleadingly has been called "streaming" --
well yes, enabling sequential processing is a prerequisite to enabling
streaming, but streaming is not the only reason for wanting to enable
sequential processing).

But of course this should not detract from the usefulness of the format
for ingest-bang implementations.
I think the structure in -02 is quite close to meeting both objectives.
I agree that the recent changes have led us away from that.

(I must admit I don't care as much about a complete analogy between XML
and CBOR formats; but as we now have found someone who cares about the
XML variant, we might be able to fix that.)

Grüße, Carsten