Re: [core] Link target attributes in CoRAL (was: Review of CoRAL)

[Klaus Hartke <[e-mail]>]

Christian Amsüss wrote:
> Only defining the mapping for known target attributes would neatly get
> us rid of the issue of link attributes. Implementations could still be
> as simple as possible (ie. not even have an enumeration of known
> attributes) if they accept producing odd output on unspecified input,
> and follow the most common rule (map to attr:${key} "${value}") on all
> but the known whitespacers.

👍

> My expectation though is that processors that convert link-format to
> CoRAL would be less constrained devices (C2 or larger), and that smaller
> devices that still want to serve both would have something equivalent to
> hard-copies of both representation in ROM instead -- is that a
> reasonable one?

As long as a constrained device understands all attributes in its
</.well-known/core>, I see not problem with a constrained device
generating a CoRAL representation from a Link Format representation on
the fly (or vice versa) (or both from a more efficient in-memory
representation). If that's not the case, then storing hard-copies of
both representations in ROM sounds reasonable to me.

Klaus