Update index server API

[enkore]

(A specification update will also be prepared)

Gist:

• Search semantics changed

  • Instead of ANDing the terms, they are ORed.
  • Many terms can be specified (email=...&email=....&phone=...)
  • JSON POST is now allowed:

   {"query": [
      {"field": "email", "value": "abc@example.net"},
      ...
    ]}

  • JSON POST is recommended, because the request line length, which contains the terms if GET is used, is limited by client libraries, HTTP frontends, application servers etc. to fairly low values (expect 4K) -- which is not the case for a JSON POST.
  • Existing code should continue to work
  • Results now contain what matched each identity:

  {"identities": [
       {"public_key": ..........., "matches": [
            {"field": "email", "value": "abc@example.net"},
            ...
       ]}
  ]}

• Client authorization will be required on official servers
  • Exact same mechanism as the block server uses.
• Phone numbers are now normalized. This shouldn't change anything for well-behaved clients that already normalize everything to E.164.

[enkore]

docker image updated. The latest-image doesn't require authorization yet, qabel/infrastructure:reqauth does.

[thechauffeur]

What is the status of this? Has the mentioned gist already been implemented? Have we decided yet on ORing?

[enkore]

As far as I'm aware nothing has been implemented yet on the core/client side of things.

What's in the gist is in the servers.