On Tue, Sep 30, 2014 at 11:23 PM, Randy Presuhn <
[email protected]> wrote:

> Hi -
>
> >From: "Y. Richard Yang" <[email protected]>
> >Sent: Sep 30, 2014 2:48 PM
> >To: Andy Bierman <[email protected]>
> >Cc: IETF ALTO <[email protected]>, [email protected]
> >Subject: Re: [netmod] Key-value data structures in YANG/JSON encoding
> ...
> >more general issue regarding completeness of data types in YANG:
> >
> >- list is an ordered array
> >
> >- container is a static, heterogeneous associate array
> ...
>
> Why do you say "static"?
>
>
It is my relative aspect, comparing list and container. The additional
substatements to list (key, max-, min-elements, ordered-by, and unique)
appear to me to enable more dynamic operations (the introduction of key
allows insert first, last, before, after) and provide more consistency
during and post dynamics (max-, min- check size, unique check duplicate,
resort if ordered by). As a contrast, container is more like a (static)
class definition in C++/java to provide simpler bundling. Of course, it
still supports basic edit-config operations, and hence in this sense is not
static at all.

I am curious on if there were discussions during the design on introducing
these two different types of interior nodes.

In the context of key-value store example, we need the key substatement to
indicate the key, and hence need to use list, but then need to implement
the additional APIs such as insert first/last/before/after which have no
meaning for a key value store. In other words,  the key-value store appears
to be a setting that may not need the helping-with-dynamic part.

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

Reply via email to