Re: [domainrep] I-D Action: draft-ietf-repute-media-type-03.txt

[Guangqing Deng <[e-mail]>]

Maybe it is worthy to take consideration of RATER-AUTHENTICITY whose value
range is also between 0.0 and 1.0. Users will be puzzled by a
RATER-AUTHENTICITY of 0.5, for they can’t determine the RATED is genuine or
not. It is better to send a clear indication to the users no matter what
they will do (but the recommendation is provided). So the
RATER-AUTHENTICITY item should also be either 0 or 1, no other
floating-point number between 0 and 1. Another issue that I am interested
in is how to compute the RATER-AUTHENTICITY? You know, the RATING is
computed from the collected data; is the RATER-AUTHENTICITY also computed
from that? Please forgive me if I miss something important.

2012/7/19 Steve Allam <[e-mail]>

>
> It depends.
>
> If my assertion is "does this sender ALWAYS send good messages", the
> rating would be 1 or 0
> If my assertion is "what percentage of mail is good from this sender", the
> rating will be somewhere between 0 and 1
>
> Hence, the rating can be whatever the application author wants it to be,
> repute just provides the framework.
>
> Steve
>
>
>
> On 19/07/2012 16:43, 邓光青 wrote:
>
> Hi, folks, before mentioning the RATER, RATED and RATING, let’s look back:
> what should the judge do towards the suspect in court. Usually, the judge
> should sentence the suspect to be either guilty or innocent; no judge
> will sentence a suspect to be 50% guilty (or innocent)! To some extent, the
> role of RATER, RATED and RATING is similar to that of the judge, suspect
> and verdict, respectively. In my opinion, the RATER should have a clear
> position towards the RATED; so the RATING should be either 0 or 1, no
> other floating-point number between 0 and 1.
>
>
> 2012/7/19 Steve Allam <[e-mail]>
>
>>  Hi,
>>
>> Maybe also worth pointing out that the reputation application that is
>> being used (ficticious in this case) would have its own description of how
>> it used the rating item, and how the number should be interpreted - in a
>> similar way to many anti-spam systems that provide a score, they normally
>> provide information on what different levels of that score mean and how you
>> might use them - as the user, you can then do whatever you like, but the
>> recommendations are provided.
>>
>> As a further example, in our reputation application, the RATING item in
>> the repute response will be either 0 or 1, never anything inbetween, as we
>> are simply returning a true or false - our documentation will detail how
>> you may wish to use the 0 or 1 as well.
>>
>> Regards,
>>
>> Steve
>>
>>
>>
>>
>> On 18/07/2012 18:37, Murray S. Kucherawy wrote:
>>
>>  On Wed, Jul 18, 2012 at 6:29 AM, 邓光青 <[e-mail]> wrote:
>>
>>>
>>>   Consider this case:
>>>
>>> Content-type: application/reputon+json
>>>
>>> {
>>>
>>> "reputon":
>>>
>>> {
>>>
>>> "rater": "RatingsRUs.example.com",
>>>
>>> "rater-authenticity": 1.0,
>>>
>>> "assertion": "IS-GOOD",
>>>
>>> "rated": "Alex Rodriguez",
>>>
>>> "rating": 0.5,
>>>
>>> "sample-size": 50000
>>>
>>> }
>>>
>>> }
>>> …indicates that we are absolutely *sure* (1.0) that the entity "
>>> RatingsRUs.example.com" consolidated 50000 data points (perhaps from
>>> everyone in Yankee Stadium) and concluded that Alex Rodriguez may be *good*
>>> or be *bad* (0.5) at something. Besides, if "rater-authenticity" equals 0.5
>>> but "rating" equals 1.0, it is also very hard to understand.
>>>
>>
>>
>> I interpret that reply to mean, literally: RatingsRUs.example.comcollected 50,000 data points about Alex Rodriguez.  It is 100% certain that
>> "Alex Rodriguez" refers to a real thing.  The data collected indicates that
>> the claim "Alex Rodriguez is good" is 50% true.
>>
>> As a person reading this, I take that to mean he's not good but also not
>> bad at whatever the application space covers, or that it's 50% likely that
>> he's good.
>>
>> -MSK
>>
>>
>>   _______________________________________________
>> domainrep mailing [e-mail]://www.ietf.org/mailman/listinfo/domainrep
>>
>>
>>
>>
>> _______________________________________________
>> domainrep mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/domainrep
>>
>>
>
>
> --
> Guangqing Deng
>
>
>


-- 
Guangqing Deng