Hi Hans,

Thanks a lot for your feedback. Please find answers to your questions inline. 
Don't hesitate to get back with further questions as it may help for the next 
draft version. 

Thanks,
Sabine


>>-----Message d'origine-----
>>De : alto [mailto:[email protected]] De la part de Hans Seidel
>>Envoyé : lundi 20 juin 2016 16:35
>>À : Randriamasy, Sabine (Nokia - FR); Kai GAO; Qiao Fu; [email protected]
>>Cc : [email protected]
>>Objet : Re: [alto] TR: I-D Action: draft-ietf-alto-multi-cost-02.txt
>>
>>Sabine, Wendy,
>>
>>Below, my comments/questions on draft-ietf-alto-multi-cost-02.txt.
>>
>>
>>S4.1.2 "constraints" paragraph:
>>
>>If I get this correct, in case "max-cost-types" > 0 and extended
>>predicates are used, the logical OR applies instead of logical AND.
[SR     ] 
[SR     ] It is perfectly fine for a MC(Multi-Cost) capable client to specify a 
"constraint" parameter composed by an array of extended predicates related by 
an AND.
The "constraints" description just stresses that the client cannot specify both 
"constraints" and "or-constraints". The "or-constraints" specification supports 
"constraints". For example:

a client willing to express "b in [12, 20] and r in [1, 10]" using 
"constraints" formulates it as follows, (assuming [0], [1] are the indexes of 
metrics b and r:

  ["[0] gt 12", "[0] le 20", "[1] gt 10", "[1] le 10"] 
  

A client willing to express "b in [12, 20] and r in [1, 10]" using 
"or-constraints" adds 2 brackets and formulates it as follows:
[
  ["[0] gt 12", "[0] le 20", "[1] gt 10", "[1] le 10"] 
]  

A client willing to express "(b > 12 and b < 20) or r < 10 or h < 2"  
formulates the or-constraint as follows:
[
  ["[0] gt 12","[0] le 20"],
  ["[1] le 10],
  ["[2] le 20"]
]  

>>Why changing the behavior? Is there an advantage in using the logical
>>OR?
[SR     ] some MC clients may find it simpler to only use one type of extended 
constraints and thus "or-constraints" supporting both "and- only" and "and-or" 
combinations is an advantage. 

>>
>>What is used if no extended predicates are used, for example when a
>>legacy client is requesting a multi-cost resource?
[SR     ] a legacy client is not allowed to request a multi-cost resource and 
must comply to RFC7285.

>>I assume the logical AND as stated in RFC7285 but it is not entirely
>>clear to me.
[SR     ] right, a legacy client MUST use the "constraints" member as specified 
in RFC7285

>>
>>How can a server distinguish between RFC7285 predicates and the new
>>extended predicates?
[SR     ] a legacy server will only understand legacy client requests. A MC 
Server MUST be able to understand both MC and legacy clients. 

>>Here, I am especially thinking of the case, that cost type index "0"
>>can be omitted and the extended predicate becomes indistinguishable
>>from
>>RFC7285 predicates.
[SR     ] a MC client must be able to talk with both MC Servers and legacy 
Servers. It MUST omit index "0" when interacting with a legacy server. It MAY 
omit it when requesting a single cost resource with constraints on that cost to 
a MC server. 
>>
>>
>>In general, I noticed some upper/lower case inconsistencies for certain
>>nouns on several occasions. E.g. "client" and "Client" or "endpoint"
>>and "Endpoint".
>>For consistency reasons only one notation should be used.
[SR     ] thanks a lot for notifying this. Will be corrected.

>>
>>
>>Regards
>>Hans
[SR     ] Thanks, Sabine
>>
>>
>>On 14.06.2016 11:30, Randriamasy, Sabine (Nokia - FR) wrote:
>>> Hi Kai, Qiao, Fred and all,
>>>
>>> We have posted a new version of Multi-Cost ALTO according to your
>>feedback. For short:
>>>
>>> - The extensions overview on "testable-cost-type-names",
>>"constraints"
>>> and "or-constraints" has been further clarified with additional
>>> subsections see 3.6.1 to 3.6.5,
>>>
>>> - the design for the testable-cost-type-names capability (section
>>> 4.1.1) has been updated according to Kai's feedback:
>>> "testable-cost-type-names" and "cost-constraints" are now mutually
>>> exclusive to prevent legacy clients from issuing constraint tests on
>>> untestable cost types,
>>>
>>> - the text on "constraints" and "or-constraints" input members in
>>> section 4.1.2 has been updated accordingly,
>>>
>>> - the or-constraint member has been corrected to [JSONString or-
>>constraints<0..*><0..*>;],
>>>    please note that this member will be corrected to [JSONString or-
>>constraints<1..*><1..*>;] in the next draft version.
>>>
>>> - the ipv4 examples in section 5.6 have been extended with ipv6
>>> examples,
>>>
>>> Again thank you all for your valuable comments.
>>>
>>> Please take a look and let us know if these updates meet your
>>> expectations, Thanks,
>>>
>>> Sabine, Wendy
>>>
>>> _______________________________________________
>>> alto mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/alto
>>
>>_______________________________________________
>>alto mailing list
>>[email protected]
>>https://www.ietf.org/mailman/listinfo/alto

_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to