[Cbor] Donation of short arcs 1.3.179.x Re: Prefix compression (Re: Review draft-bormann-cbor-tags-oid-07)

[Sean Leonard <[e-mail]>]

On 7/4/2020 4:48 AM, Carsten Bormann wrote:
> Hi Jim,
>
> On 2020-07-04, at 03:53, Jim Schaad <[e-mail]> wrote:
>> An array of two hash functions which could be parsed into the two rooted OID values
> Prefix compression is applicable to other data than OIDs (most prominently URIs).
> I think we should handle that in a more generic way than specifically for OIDs.

I control a very short OID arc, 1.3.179 {iso(1) 
identified-organization(3) penango(179)} and would be happy to donate a 
block of 16 one-byte identifiers for CBOR/CBOR working group (or IETF) 
use. The binary encoding of this prefix is h'2B8133'.

If I assign the block of 96-111 to CBOR (h'60' through h'6F'), then that 
yields four bytes, 2048 unique numbers that are just five bytes long and 
16 unique numbers that are four bytes long as a bonus. I could probably 
donate 32 one-byte identifiers without losing sleep over it.

OIDs are just a way to generate globally unique, language-agnostic 
identifiers. A unique identifier that is five bytes long is clearly 
superior to a web URI (which depends on perpetual domain name 
registration) or even a UUID (fixed 16 bytes long), and obviates any 
need for compression. I think the reason why OIDs get such a bad wrap is 
because they are unnecessarily long when politics are laid over them, 
like 1.3.6.1.4.1.9.9.179.1.3.3.1.13. If you want a short and globally 
unique one, you can't get better than OIDs.

Notably, I am not the only one to figure out how to get a short OID: 
looks like Amazon picked up #187 in 2015: 
<http://oid-info.com/get/1.3.187>. If you can convince anybody who has 
registered OIDs 1-127 to slice off a piece of their namespace, you can 
make them one byte shorter. And I just discovered that IANA itself has a 
registration under 1.3.90: <http://oid-info.com/get/1.3.90> which is 
h'2B5A'--just two bytes! I'm surprised that nobody has made use of it.

While there are a lot of approaches to prefix compression, this assures 
that you can use OIDs instead of other identifiers in CBOR for *globally 
unique* identifiers. It has another modern advantage in that 
www.oid-info.com is a publicly accessible wiki-style list of all the 
OIDs ever minted, so you can write a tool to retrieve the human-readable 
meaning of any OID with a simple HTTP GET request.

Sean