Re: [Insipid] MUST or SHOULD for Type 4 / 5 UUIDs?

["Paul Giralt (pgiralt)" <[e-mail]>]

I don’t see how having a MUST causes a backward compatibility issue. At the end of the day, a UUID is just a string of 32 characters and as long as that criteria is met, it really doesn’t matter how the string was generated. Any device following this draft would generate a Type 4 or 5 UUID and a pre-standard device might generate it in some other way and everything will work fine. We are specifically addressing how a behavior implementing this draft generates a UUID. An endpoint could end up using a non Type 4/5 UUID as a remote-uuid when interoperating with a pre-standard device, but the device following the standard is not generating the UUID in that case, so it is not violating the MUST clause in that case.

-Paul


> On Jun 2, 2016, at 3:01 AM, Christer Holmberg <[e-mail]> wrote:
> 
> Hi,
> 
> Is there a reason we put SHOULD in section 4? Backward compatibility?
> 
> Regards,
> 
> Christer
> 
> 
> On 02/06/16 07:00, "insipid on behalf of Paul Giralt"
> <[e-mail] on behalf of [e-mail]> wrote:
> 
>> INSIPID WG,
>> 
>> Ben pointed out that we have an inconsistency in the draft. Section 4
>> states that endpoints SHOULD use Type 4 or Type 5 UUIDs when generating a
>> Session Identifier, however Section 12 states that Type 4 or 5 MUST be
>> used for privacy reasons.
>> 
>> I¹m inclined to change the first to MUST and thereby require that only
>> Type 4 or Type 5 UUID¹s be used.
>> 
>> Does anyone have an issue with this change?
>> 
>> -Paul
>> 
>> _______________________________________________
>> insipid mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/insipid
>