Re: [core] [Anima] documenting SID usage in IETF specification

[Andy Bierman <[e-mail]>]

On Thu, Sep 13, 2018 at 8:18 AM, Michael Richardson <[e-mail]>
wrote:

>
> {removed anima from CC list}
>
> Carsten Bormann <[e-mail]> wrote:
>     > What Andy pointed out is that we also need to have an idea of how to
>     > evolve a draft in a way that minimizes damage from changing those
>     > numbers during the development of that draft.  So we need to start
>     > allocating and managing SID numbers early in their lifetime, at least
>     > from the point of time when a draft is becoming an “implementation
>     > draft” (as opposed to just an idea that wants to be discussed).  That
>
> Agreed.
> I think that we will typically be need about ~100 per draft, sometimes as
> many as 1000, but never 10,000.  100 is a very small number compared to the
> size of the SID space.
>
> I think that the early allocation process will work fine, once we have a
> registry to allocate against.  The comi.space mechanism is even lighter
> weight.
>
>     > — have a more explicit way of designating drafts as Implementation
>     > Drafts.  Basically, any SIDs allocated before that are without
>     > protection, but once we have an Implementation Draft, the SIDs used
> in
>     > that will not be re-used.  (Intermediate versions between
>     > Implementation Drafts would again have any new SIDs in unprotected
>     > state until another Implementation Draft is declared.)
>
> I think that we used to call these "Proposed standards" :-)
> I'm not sure we need to have this designation.  I think that SID numbers
> are
> cheap enough that it's okay if some drafts consume some and then get
> abandonned.   The numbers could be recycled 6 months after the draft
> expires.
>
>

IMO you had it right up to the last sentence.
There are not port numbers. There are 2^^64 of them.
Recycling SID numbers would be a recipe for disaster.
The "experimental" SID space is especially problematic in this regard.
Within a deployment there will be server code from all different dates,
so being able to declare the SID space clean at 1 single point in time is
not good enough.

For "textual YANG" this is not a problem because there is very low
probability
that module names or module namespace URIs will collide.




>     > — have a way to include the SID file in the document (draft, RFC).
>     > This is not beautiful, but unless we invent another representation
> for
>     > that information, that is the interchangeable form.  (If we do invent
>     > another representation, maybe we should always use that?)
>
> I suggest it be included as an appendix.
>
> --
> Michael Richardson <[e-mail]>, Sandelman Software Works
>  -= IPv6 IoT consulting =-
>
>
>
>
Andy