Re: [Sedate] Can offsets like [+02:00] be used instead of IANA names in brackets?

[Justin Grant <[e-mail]>]

The purpose of the time zone extension is to allow senders to tell
receivers about the time zone of the timestamp. A timestamp can only have
one time zone, so therefore there should only be one time zone extension.

An IANA name in brackets means "receiver, please use your TZDB rules to
decide how to calculate the offset of derived timestamps like "1 day later"
An offset string in brackets means "receiver, I am opting out of using time
zone rules. Instead, every derived timestamp should use this offset."

A good question is why not just use a bracketless RFC 3339 string to
indicate opting out of using TZDB rules?  The main reason is that it'd
unnecessarily break compatibility with existing Java code.

But there's another reason too. Programs that create DST-aware derived
values (e.g. "one day later") from an input string should require bracketed
timezone annotations and therefore should fail parsing if the annotation is
missing. Doing this ensures that senders won't inadvertently pass an RFC
3339 timestamp (with no awareness of its time zone) to a program that needs
the time zone to calculate derived values.  But if the sender wants to opt
out of time zone processing, then using an extension like [-08:00] allows
that escape hatch without opening up the receiving program to bugs caused
by senders accidentally omitting the time zone annotation from the string.

Note that offset time zone annotations won't be a common case. It's an
escape hatch for cases where the time zone is not known, and when the
sender wants to tell the receiver that they're OK with it being unknown.
Had Java not accepted this syntax we probably wouldn't add it, but given
that Java does support this syntax and it's been in wide use for 10+ years,
it seems prudent to accept it in the standard.

Cheers,
Justin

On Tue, Mar 8, 2022 at 12:56 PM Ujjwal Sharma <[e-mail]> wrote:

> I believe the Java format only allows one bracketed suffix, but also it
> won't be quite useful to have two conflicting pieces of information.
>
> Why *did* Java support offsets in the brackets, however, is not known to
> me. Perhaps we could ask Stephen Colebourne
> (https://twitter.com/jodastephen)?
>
> Ujjwal
>
> On 09/03/2022 01.50, Carsten Bormann wrote:
> > On 2022-03-08, at 21:19, Justin Grant <[e-mail]>
> wrote:
> >>
> >> IMO, "second exception to the [foo=bar] syntax" sounds like
> [-08:00][America/Los_Angeles] could be allowed in a string, when in reality
> the time zone suffix should be mutually exclusive: either an IANA name or a
> numeric offset string.
> >
> > Ah, that’s the part I missed.
> > So what is the reason these are mutually exclusive?
> >
> > Grüße, Carsten
> >
> >
>
> --
> Ujjwal "Ryzokuken" Sharma (he/him)
>
> Compilers Hacker, Node.js Core Collaborator and Speaker
>