Re: [Cbor] Interactions of packed CBOR and tags

[Brendan Moran <[e-mail]>]

> On 3 Sep 2020, at 20:55, Carsten Bormann <[e-mail]> wrote:
>
> On 2020-09-03, at 18:57, Brendan Moran <[e-mail]> wrote:
>>
>> However it’s still going to be bigger than a cbor-encoded plain text URIs if you have more than a few > 23 byte path segments, which is a bit annoying.
>
> It’s all in your head :-)
>
> The point of CBOR (which also was the point of JSON) was to be a reasonably appropriate encoding for much of what an application wants to do (the difference between JSON and CBOR is in what is “reasonably appropriate” in the areas of application).
>
> There is *always* a way to create a more compact bespoke encoding.  For anything.
> For URIs: We haven’t even started Huffman-coding the string parts.  (Or LZ-coding with some smart predefined dictionary.  Whatever.)  So why worry about the length indicator for > 23 byte path segments, where the text itself could be so much more efficiently coded?
>
> Engineers have a tendency to react to perceived optimization opportunities.  That is of course a good thing.  It turns bad when we start to tweak something and incur complexity costs just because it is slightly worse than something else.
> (If it is much worse in some metric, and that particular metric matters, well then back to the drawing board.  But not here.)
>
> Learning to use CBOR includes learning to tolerate the occasional bouts of cognitive dissonance that come from knowing “I could do this better” where the “better” is a few percent.

Normally, I’d agree with you. However, I did carefully consider the tradeoffs in this case. My argument is that there are two competing optimisations here:

1. Reduce parsing overhead for handling URIs by encoding them in CBOR
2. (Simply) Reduce encoding size of URIs.

I agree that scheme identification is better done by enum. I agree that domain name/IP is better identified with explicit text for domain name or bytes for IP. I agree that port number is better encoded as an integer.

I’m not yet convinced that there is a substantial reduction in complexity to be had by encoding a textual path as an array of CBOR text elements versus ‘/‘ delimited elements. I’ll have to go read through the mail archives on that one.

If it turns out that there is not a substantial reduction in complexity, then I don’t see why it would make sense to encode them in this way, rather than sticking with conventional URI path encoding.

Brendan
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.