Re: [Cbor] Self Described CBOR - Little Endian

[Felipe Gasper <[e-mail]>]

> On Mar 23, 2021, at 12:50 AM, alex thompson <[e-mail]> wrote:
> 
> ​Given the prevalence of little endian hardware and the existence of the LE typed array tags it would be useful to opt-in to LE encoding of the additional information bytes as well. This would simplify decoders that are specific to LE platforms, particularly WebAssembly which is all LE and hasn’t defined a swap instruction.
> 
> The opt-in could be self described with tag 55798 (has the same non-unicode magic number characteristic as 55799). The beginning of a stream would default to BE so tag 55798 itself would appear as 0xd9d9f6. Then any following tags and descendant items would be LE.
> 
> For completeness, 55798 and 55799 could be nested within one another, switching the endianness as the decoder traverses a tree of items. Nested tags 24 and 63 would pass on their endianness to their embedded items. No effect on typed array tags 64-87.
> 
> It would be a breaking change for generic decoders since ignoring 55798 wouldn’t work as expected but I think the impact is very manageable and fits with CBOR’s stated goals of small code size and extensibility.

I, too, have lamented the use of big-endian byte order in CBOR, FWIW. I *think* my Perl generic decoder works on big-endian machines, but I don’t think I have access to any big-endian machines to actually _test_ it.

-F