Re: [alto] Review on draft-ietf-alto-unified-props-new-04

[Jensen Zhang <[e-mail]>]

Hi Kai,

Thanks for your review. Please see my comments inline.

On Sun, Dec 9, 2018, 10:25 AM Kai GAO <[e-mail]> wrote:

> Hi everyone,
>
> Below is a review on the unified property map extension:
>
> 1. In Sec 2.1, the first sentence reads "The entity is an extended concept
> of the endpoint ...". Here the word "extended" may not be very precise, and
> the term "generalization" (which is also used in the abstract) sounds
> better. Generalization indicates that an endpoint is essentially an entity
> while extension could be misleading and even incorrect. For example, in
> certain languages, A extends B indicates that A is also B.
>

The term "generalization" sounds more precise. If nobody has strong opinion
to disagree, I will update it in the next revision.


> 2. In Sec 2.2, an entity domain is defined as "a set of entities". This
> seems odd because then one can say a set of two entities
> {"ipv4:190.0.2.34", "pid:PID1"} is also a domain, which doesn't make sense.
> An entity domain should be a generalization of endpoint address type, which
> must define the syntax and semantics of the entity addresses in this
> domain. Thus, borrowed from the definition of a domain in math, it could be
> "the complete set of all possible values of a given address type". Here the
> "given address type" is uniquely represented by the domain name, which
> indicates the "semantics" for this domain, while syntax for "all possible
> values" is defined by the "domain-specific entity addresses".
>

Exactly. I will change the wording.


> I also feel Sec 2 can be slightly rearranged for better clarity. Right now
> there are a lot of cross-references between different concepts. I suggest
> having a short section introducing the terms and then using a paragraph to
> specify their relations, for example,
>
> (domain name, domain-specific address type, hierarchies, relations)
> -(1:1)-> domain -(1:n)-> entity address -(1:1)-> entity -(1:m)-> property
> <-(1:1)- (property type, property value)
>

Good point. I will think about it.


> 3. I think the draft should make it clear that the uniqueness of an entity
> address only applies in the same unified property map. For example,
> "pid:PID1" could point to different entities when two UPMs depend on two
> different network maps, both have the PID "PID1".
>

I agree. I will make it clear.


> Best,
> Kai
> _______________________________________________
> alto mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/alto


All the comments on the definition part make the document clearer. I will
update them.

Thank you so much!
Jensen