Re: [Cbor] Interactions of packed CBOR and tags

[Carsten Bormann <[e-mail]>]

On 2020-09-03, at 18:03, Jim Schaad <[e-mail]> wrote:
> 
> 6([ "www.", "merch", "datatracker"], [".ietf.org""], 224(225(simple(2))), 224(226(simple(2)))] 
> 
> Where we have pulled both prefix and postfix strings extracted and maximize the amount of commonality.  

I like the idea of adding suffix packing.

In the current proposal, to distinguish shared items from prefix packing, we have

— a separate table (and thus number space), which makes sense if prefixes are rarely used as shared items and v.v. (using a prefix as shared item just requires prefixing it to the empty item, though);
— separate referents (one with no content, one with the content the prefix is prefixed to).

Adding suffix packing, we could have a third table and a third set of referents (also with content).  We also could have just separate referents, sharing the table/number space, if that makes sense.

We could also rethink the whole separation of shared items from prefix/suffix items in this process.

As Brendan says, being able to optimize the tables so the frequently used items land in the areas that have short referents is important.

[Apologies for using the term referent, which isn’t even in my dictionaries with the meaning I’m using it for (the thing that references).  But that’s the usage I learned when I learned about them in the 1970s… [1]]

Making prefix/suffix packing well-defined for arrays isn’t hard.  For maps, there is no difference (assuming we are always handling full key/value pairs — just doing a single key or a value can be done by sharing items).

I don’t think we want to address what would be “deterministic packing”.

Grüße, Carsten

[1]: https://scholar.google.com/scholar?hl=en&q=Ross%2C+D.+T.%2C+%22Uniform+Referents%3A+An+Essential+Property+for+a+Software+Engineering+Language.%22+Paper+presented+at+the+3rd+International+Symposium+on+Computer+and+Information+Sciences+%28COINS-69%29%2C+Miami+Beach%2C+Florida%2C+18%2D%2D20+December+1969.