[Cbor] Re: Consensus call on EDN literals single ABNF

[Pete Resnick <[e-mail]>]

On 25 Jul 2024, at 11:58, Carsten Bormann wrote:

> On 24. Jul 2024, at 01:30, Pete Resnick 
> <resnick=[e-mail]> wrote:
>>
>> app-string = app-string-h
>> / app-string-b64
>> / app-string-dt
>> / app-string-ip
>> / app-string-gen
>>
>> app-string-gen = app-prefix sqstr ; generic app-string
>
> So why don’t we always use app-string-gen?

To a certain extent, I don't know what you mean by "always use 
app-string-gen". Do you mean that you want the parser to always result 
in an app-string-gen token? That's fine, but as I said, that means that 
the result of the parse will never end up with a bstr containing an 
app-string-ip, for example.

> (That is the two-layer model.)

ABNF does not have layers. You are of course free to implement a 
multi-layer parser and implement the parsing of "app-string" separately 
from other parts of the grammar, but the ABNF language doesn't have a 
way to indicate that.

> The way you write it looks like a premature optimization for the five 
> cases that happen to be in this document.

ABNF productions are extensible. Currently, the only 5 cases in the 
grammar are the ones that the document describes. You want to add more, 
you can simply "=/" another case.

The way the ABNF is currently in the document is simply incorrect 
AFAICT.

pr
-- 
Pete Resnick https://www.episteme.net/
All connections to the world are tenuous at best