Re: [calsify] AD review of draft-ietf-calext-valarm-extensions-03

[Ken Murchison <[e-mail]>]

On 1/29/21 1:02 PM, Barry Leiba wrote:
> Ken, I think your interpretation is correct.
>
> One way to handle the various combinations of required "things",
> optional-but-once things, and optional-multiple things is something
> like this:
>
>     abc = req opt-once *("," opt-mult)
>
>     req = r1 "," r2 "," r3
>
>     opt-once = ["," o1] ["," o2] ["," o3]
>
>     opt-multi = m1 / m2 / m3 / r3
>
> In this case, r3 is required once and can appear multiple times, so
> it's in both places.
>
> The problem with that mechanism is that it's hard to extend anything
> except opt-multi.  You can do
>
>     opt-multi =/ m4
>
> ...but you can't do anything similar with req or opt-once.  And also,
> req and opt-once are ordered, which is sad.
>
> I'm chatting with Dave Crocker and some others about whether it makes
> sense to extend ABNF to support unordered lists.
>
> Anyway, the more we talk about this, the more convinced I am that
> Cyrus is right: we should leave the ABNF alone here, and deal with
> properly updating it in an update of the base doc, if we ever get to
> that.


Agreed that we shouldn't try to reinvent the wheel here, but still I 
think our current use of alarmprop =/ foo in this draft follows previous 
use, but doesn't actually do what we want.

I also think this is an issue in 
https://tools.ietf.org/html/draft-ietf-calext-eventpub-extensions-18#section-4 
<https://tools.ietf.org/html/draft-ietf-calext-eventpub-extensions-18#section-4>


-- 
Kenneth Murchison
Senior Software Developer
Fastmail US LLC