Re: #1028 USEFOR 3.1.2 Date: Resolved, I think.

[Russ Allbery <rra@xxxxxxxxxxxx>]

Frank Ellermann <nobody@xxxxxxxxxxxxxxxxx> writes:
> Russ Allbery wrote:

>> I'll argue in favor of that, since otherwise we would be going from
>> blessing use if reluctantly (RFC 1036) to allowing parsers to disregard
>> them without any intervening deprecation period.

> Sigh.  That's the point of this thread.  It's a general design principle
> of USEFOR to replace 2822 "MUST accept obs-ceneties" by a "MAY accept"
> keeping the 2822 "MUST NOT generate".

I've clearly missed the point of the thread, and I'm sorry about that.  I
was previously responding to the language about GMT and was also looking
primarily at the generate side.  I had not realized that people were
talking about not accepting the North American time zones.

The military time zones were deprecated by RFC 1123, and more tellingly,
they also don't appear in practice.  I have no problem getting rid of them
except insofar as it adds wording complexity.  But I *do* have a problem
going straight from RFC 1036 to allowing servers to not even parse the
North American time zones, which are still in use (albeit not heavily).

I think we need one more step between here and there.

> That's exactly what we want.  And otherwise "you MAY parse it, else you
> SHOULD treat it as unknown -0000" is fine.

No, I don't agree that this is fine.

> That's in Bruce's obs-package, we're cherry picking, and IMHO reducing
> this to the one cherry GMT is an excellent choice for new parsers.

I also don't agree with this.

> Really, can we close this #1028 now ?

Apparently not.

-- 
Russ Allbery (rra@xxxxxxxxxxxx)             <http://www.eyrie.org/~eagle/>