Re: [Slim] I-D Action: draft-ietf-slim-negotiating-human-language-07.txt

[Gunnar Hellström <[e-mail]>]

Den 2017-03-01 kl. 18:47, skrev Randall Gellens:
> At 8:26 AM +0100 3/1/17, Gunnar Hellström wrote:
>
>>  Hi Randall,
>>
>>  Den 2017-03-01 kl. 02:08, skrev Randall Gellens:
>>
>>>  Hi Gunnar,
>>>
>>>  I'm starting a new message to cut out the huge amount of quoting.
>>>
>>>  Your proposal is that text be added that advises the calling client 
>>> to place an asterisk on the least-preferred language/media, and 
>>> advises the answering client to indicate to the answering human 
>>> which language/media is not the least preferred (did not have an 
>>> asterisk in the offer), is that accurate?
>>>
>>  Yes, with slight rewording to:
>>  "text that advises the offering client to place an asterisk on the 
>> least-preferred language/media indications, and advises the answering 
>> client to indicate to the answering human which language/media are 
>> not the least preferred (did not have an asterisk in the offer)"
>>
>>  The inclusion of the "indications" is just to assure that it is 
>> clear that it does not need to be just one indication that gets the 
>> asterisk .
>>  The last part sounds awkward, but matches technically what the lack 
>> of an asterisk means. I inherited the inverted logic for the asterisk 
>> from its already defined non-denial meaning.
>>  If you are considering wording for the draft, I suggest that you 
>> straighten the logic to say "which language/media are most preferred 
>> (did not have an asterisk in the offer)"
>>
>>  It does also not need to be an "answering human" that gets this 
>> indication and makes use of it for guidance on how to answer the 
>> call. It can just as well be e.g. a multi-modal answering machine or 
>> some other application interacting with human language. I am not sure 
>> if "answering party" is more appropriate and can be considered 
>> including such automata.
>
> Hi Gunnar,
>
> Thanks for clarifying, I think I understand your proposal in detail 
> now.  After thinking it over, I still think this would be better done 
> in a new draft, because (a) it is advice on a way of using the 
> mechanism to convey additional information; (b) it would be good for 
> the group to discuss the proposal and work through various cases 
> (e.g., what if the offering client is not going to include an 
> asterisk, what if there is more than one most-preferred language); and 
> (c) it would be good for the group to decide if this meets your need.
Randall,
Good that you understand it now.
I realize that this kind of added rules for an already existing 
parameter could be specified in an additional draft. Especially since it 
has no impact on the current meaning of the asterisk.
I still think it is best to add the few words needed now. The preference 
indication is so severely unbalanced without it, in that only preference 
between languages in the same modality can be specified. I am afraid 
that it can be seen as a discrimination against those who would need to 
specify preference between different modalities in order to tget equal 
opportunities to get smoothly performed calls through, but cannot.

You are right that there are situations that will not be explained if we 
accept my little extra sentence or something similar.
That is true also for the currently specified indications. We have said 
that we nearly only specify the indications and not how the negotiation 
shall be performed. It can be a topic for a BCP to advice on how the 
negotiation could be performed both with the currently specified 
language and in-media preferences and with the additional preference 
between media. We could discuss e.g. the case when two same spoken 
languages are specified but with the opposite preference order by the 
offeror and answering party. A decision must be taken, because the 
protocol says that oly one language per media and direction may be 
indicated in the answer, and also that the answering instance need to 
become aware of which language it shall produce.
I do not say that we need to resolve this case. It can be discussed in a 
BCP, and indicated that additional policy may be applied for solving 
that kind of undefined cases.

Similarily, there will be situations with the additional use of the 
asterisk that will be good to provide extra information for in a BCP. 
The preference indication is very rough, with only two levels. So there 
wil of course be situations when the users will wonder how to set their 
profile, and cases when the negotiation will be hard to assign a well 
motivated result. But we have said that we want to have the 
specification on this rough level.

The two cases that you bring up can have this treatment:

1. If the calling user want to get the call denied if no languages 
match, then the user must make a decision if that preference is more 
important to specify than the preference between modalities. In order to 
keep complexity low, I do not think that we should specify how to code 
both preferences.

2. If there are more than one most-preferred language. I understand this 
as for example a user is equally happy to use French sign language as 
spoken French, and can also, but on lower preference level write French 
text. That would be indicated by an asterisk on the French text, and the 
answering party having all these three capabilities may omit the French 
text from the answer but keep the others. Then the answering user select 
one of the spoken French and French Sign Language for its start of the 
call, knowing that the caller will be approximately equally happy with 
the call in both these cases.   Was that the case you thought about for 
the second case?

I understand that you wanted to check more than these two cases and 
discuss them with the WG, so the above is just a start. We can do more 
cases if you want, but I do not think we need any lengthy discussion 
that would delay the current draft.


>
>
> A new draft, especially one that will be either Informational or BCP, 
> can be done fairly quickly. It could be quite short, perhaps only a 
> page or two of real text plus the boilerplate text.  I am happy to 
> help with it.
>
Tanks,

Gunnar

-- 
-----------------------------------------
Gunnar Hellström
Omnitor
[e-mail]
[phone]