Re: [core] draft-ietf-core-sid-05 and draft-ietf-core-yang-cbor-07

[Robert Wilton <[e-mail]>]

On 15/03/2019 15:42, Carsten Bormann wrote:
> On Mar 15, 2019, at 14:24, Rob Wilton (rwilton) <[e-mail]> wrote:
>> In short, I think that it would be more pragmatic if SIDs are restricted to the non-negative range of the signed 64 bit integer type.  I.e. disallow values from (2^63 through 2^64 - 1).
> SIDs are unsigned integers.  However, there are still languages that have a hard time with those.

I imagine that you may be referring to languages like Java and Python, 
but in my experience, C and C++ also struggle to safely use unsigned 
numbers.

E.g. https://blog.regehr.org/archives/268 and their similar cases.


>    And it helps if all SID deltas fit into a signed 64-bit integer.

This is really the key one for me.

Otherwise, I would guess that many implementations will probably just 
assume that the delta always fits into a signed 64-bit integer and 
potentially fail silently in the future when they hit an 
overflow/underflow condition.


> So maybe limiting SIDs to the (non-negative) range of signed 64-bit integers is a restriction that makes sense.
>
> https://github.com/core-wg/yang-cbor has three pull requests, but no issues.
> (This apparently also happens to be the repo for the SID specification.)
>
> Should we put up the issues there?

I've added the author alias.  Hopefully we'll get an answer.

Thanks,
Rob


>
> Grüße, Carsten
>
> .
>