[Cbor] CDDL 2.0 import and paths

[Christian Amsüss <[e-mail]>]

Today's interim had some discussion about how bare-name CDDL files can
be imported from statically shipped libraries, the local directory, and
eventually a web-wide catalog.

I made a very incomplete suggestion of "could be URIs" to the question
of how to treat nested references; here's what I'd propose for this in a
slightly fuller form:

* Import "from"s are URI references.
* Tools should not dereference URIs by default, but may offer a `--web`
  switch to load any from HTTP(S),
  `--web=https://origin1,https://origin2` to allow web access only for
  some origins, and a `--file=ROOTPATH` flag to allow file system access
  up to ROOTPATH (wher `--file=.` is a practical default).
* Tools are encouraged to ship cache directories that are enabled by
  default. These look roughly like `wget -m` outputs in the file system,
  are used in sequence to resolve files (maybe with --web as a
  fallback). The user should be allowed to add such caches (a la
  CDDLPATH).
* An online catalog service can offer an easy to upload and quality
  controlled archive, which tools may auto-update from, possibly lazily
  (hitting an unknown URI in non--web mode, and the archive is stale?
  Better refresh from the catalog...).
* Locally used tools may have a path of bases to try when dereferencing
  as long as references are only relative. For example, when working in
  an I-D repository, there could be a fallback base that smoothes over
  the differences between the draft being hosted locally and its
  uploaded form.

I'm not sure this solves all issues, but I think it might work. To some
extent it hinges on IETF documents being assigned stable canonical URIs,
though :-/

BR
c


-- 
To use raw power is to make yourself infinitely vulnerable to greater powers.
  -- Bene Gesserit axiom