[Standards] Message-IDs

Kevin Smith kevin.smith at isode.com
Wed Feb 28 14:59:10 UTC 2018

On 28 Feb 2018, at 14:47, Denver Gingerich <denver at ossguy.com> wrote:
> On Wed, Feb 28, 2018 at 08:59:01AM +0000, Kevin Smith wrote:
>> On 13 Feb 2018, at 16:57, Simon Friedberger <simon.jabber at a-oben.org> wrote:
>>>    E3. Simply make the ID: FROM-TIMESTAMP.
>>>        Here FROM needs to be the eventual FROM after possible
>>> rewriting. Can
>>>        that be done?
>>>        And TIMESTAMP has to be strictly increasing so should have
>>> sub-second
>>>        resolution.
>>>        I assume this is impossible because otherwise it would be to
>>> easy. But
>>>        why is it impossible? :)
>> Because timestamps aren’t monotonic? :)
> Do you mean because most people use Unix time and/or other UTC-based timestamps (that have leap seconds)?
> If so, this can be mostly solved by using TAI timestamps.  Unfortunately, it is tricky in most OSes to obtain a TAI timestamp, but I found some code that does this (on many platforms anyway):
> https://ossguy.com/tai.c
> We've used this code for implementing usage tracking in JMP (to ensure a day's length doesn't vary from day to day - it is always exactly 86,400 seconds long).  For details, see https://gitlab.com/ossguy/sgx-catapult/commit/31c2cb7c8fbea1ad4cc6753a4343dbfc65552fa5 .  As you might suspect, we'd like to port the above TAI code to Ruby, but it works ok as-is for now.
> I realize that clock skew could still cause the TAI timestamp that your OS returns to be non-monotonic (i.e. a machine issue, not an issue with TAI time itself); I'm not sure if that's a substantial issue for the message IDs being discussed here.

I meant because clock skew is a thing, so relying on the monotonicity doesn’t work. Seems like it shouldn’t be a thing, but is.


More information about the Standards mailing list