On Sep 9, 2013, at 17:44, "Pierre Thierry" <pierre(_at_)nothos(_dot_)net> wrote:
Carsten,
in draft-bormann-cbor-05, you mention that "The format must be
applicable to (…) high-volume applications." in section 1.1.
Do you already have a benchmark in place to measure the volume that a
CBOR implementation is able to manage? Do you plan to compare CBOR
implementations to implementations of other similar formats?
Hi Pierre,
of course, I do have some informal benchmarks for my implementations; their
purpose is to avoid performance regressions when evolving the code.
However, it has been my experience that, unless you make mistakes in defining
the format (e.g., requiring extensive data conversions, two-pass
implementations, or back-patching) the actual performance of a data format
encoder/decoder is largely dependent on system issues, such as the data model
you are using and the memory allocation principles. So I'm not sure a formal
benchmark would help a lot.
What kind of application do you have in mind? Maybe I can come up with a
better, more tailored benchmark for that.
Grüße, Carsten