Re: [sieve] Working Group Last Call: draft-ietf-sieve... autoreply, notify-presence, vacation-seconds

[Tony Hansen <[e-mail]>]

On 10/13/2010 9:55 AM, Barry Leiba wrote:
>> =======
>> notify-presence
>>
>> The value of "unknown" should be explicity listed in section 2 for each
>> item that can return that value, rather than being mentioned in passing
>> at the end of section 2.
>>      
> I don't agree; "unknown" isn't a value of the presence, but an
> artifact of retrieving it; I think it's worth keeping that separate.
> Would you be happier (I think I would) if I moved it up?  My working
> copy now has this as the second paragraph, just before the list:
>
>     This document defines a set of items of notification presence, which
>     may be specified in the notification-capability parameter.  The
>     script tests the values of notification presence items in the key-
>     list parameter.  The values that each item may have are specified in
>     the list below; in addition, any item may have the value "unknown",
>     if it is not possible to determine the correct value of the item.
>    

I disagree. For example, under show it says:

	The value of this item is one of the following:

That certainly includes "unknown".

I think it's just easier to include

	unknown - the correct value could not be determined

in the appropriate places.


>>> I think it would be worth putting a note in somewhere that an
>>> implementation may wish to cache the results of a call to
>>> notify_method_capability so that if it is used in a series of if/else
>>> tests, it only needs to be retrieved once.
>>>        
>> Now that you bring this up, I'm tempted to go even further, and say that any
>> result should be consistent throughout the script. If statuses change and
>> notify_method_capability returns inconsistent results at different times in
>> the same script, I could see that causing some unpleasant results.
>>      
> I like Ned's suggestion, and I will add that.  My working copy now has
> this as the third paragraph, just before the list:
>
>     If a particular presence item is tested multiple times within the
>     same script execution context, implementations MUST present the same
>     value each time (for example, by caching the value on first use).
>     This provides consistency within a single execution.
>
> Please let me know if you guys are happy with these changes, and I'll
> post a revision.
>    
+1

     Tony