Re: [hybi] CML

["Shelby Moore" <[e-mail]>]

> You are conflating layers.
>
> The sending layer will serialize the Javascript data structure, perhaps
> into JSON (even if it was already a JSON string being transported), and
> then the receiving layer will de-serialize it.  The de-serializer code
> needs to know how to hop from frame fragment to next one.
>
> The JS coder never sees that.  That de-serializer is a low level operation
> inside the browser code.

And thus your current statistics on mean frames sizes could end up being
too low.

Maybe the serializer will be more efficient than JSON.

But I think it is really wrong to try to squeeze 2 - 4% of blood with the
"Option #2 - 7/16/63-bit" at the possible disaster having most frame sizes
end up greater than 126 once we get all the conflation out of the design.

Must better to add the extra byte and choose either "Option 15/63-bit" or
"Option 8/16/32/64-bit" so as to get some margin of error and safety.

Also safety against my alleged game theory disaster below, regardless
whether you consider it likely or not, it just isn't good engineering to
not allow a margin of error for safety:

http://www.ietf.org/mail-archive/web/hybi/current/msg03578.html