Re: [yang-doctors] YANG QUIC-LB model

[Martin Björklund <[e-mail]>]

Hi,


Martin Duke <[e-mail]> wrote:
> Hello yang-doctors,
> 
> I don't know that this rises to the level of a full review, but the QUIC-LB
> draft (currently adopted by the QUIC WG) is basically a standard for
> configuration of QUIC Load Balancers and servers to work together. After
> much reluctance to take on a format I don't understand, I took a shot at a
> YANG model:
> 
> https://github.com/quicwg/load-balancers/blob/yang/draft-ietf-quic-load-balancers.md
> 
> (The YANG stuff is in an appendix near the end -- search for #yang-model).
> 
> Before we talk about a full review where you labor to understand the data
> items, I'd appreciate an informal quick sanity check that I am doing this
> roughly correctly, as I have zero experience with YANG.

I downloaded the module and ran it through a YANG compiler, and
started to fix several minor syntactical errors (such as missing ";"),
but then I ran into other issues that made it difficult for me to
guess what you meant.  I suggest you ensure that the model compiles
(there are a bunch of compilers available, e.g., pyang, yanglint, ...)
and post it again.

Some comments though:

You have:

       list cid-configs {
         key "config-rotation-bits";

         leaf config-rotation-bits {
           type uint8 {
             range "0..2";
           }

This means that there can only be max 3 entries in this list.  Is that
what you meant?


You also have:

       typedef quic-lb-key {
         type yang:hex-string {
           length 16;
         }
       }

The YANG statement "length", when given on a type "string", or derived
from "string", operates on the number of characters in the string.  In
this case, you probably what to limit the number of bytes represented
by the hex-string.  Unfortunately, it is not possible in standard YANG
1.1 to do this.

However, in this particular case it sort-of works, since the length of
the hex-string is 2 * N + 1, where N is the number of bytes.  Hence,
the length would be 33.  If you do this I suggest you explain this in
the description statement.

An alternative could be to drop the length statement and only explain
it in the description statement.

> One other silly question: I have a field whose length depends on other
> fields. Can you give me an example on how to express this correctly? e.g.
> 
> leaf foo-length {
>     type uint8
> }
> 
> leaf foo {
>     type yang:hex-string {
>         length foo-length;
>     }
> }

This is possible by using a "must" statement.  However, it has the
same problem with the lexical value as above:

  leaf foo {
    type yang:hex-string;
    must 'string-length(.) = 2 * ../foo-length + 1';
  }



/martin