Search: links in ad-hoc queries

[cportele]

A follow-up to #755: Properly handle links in responses to ad-hoc queries (links imply use of GET, but ad-hoc queries use POST with JSON payload). The currently links are not correct.

The self and alternate links should simply be suppressed, there is no resource on the server that we can link to (this is a transient ad-hoc query).

With respect to the next link, if there are more matches than the page size, here are some options:

• No support paging with ad-hoc queries.
• Use an extended link object that includes the information to use the POST method and includes the request content and content type. The content is the same as from the request - with an adjusted value for offset.
• Store that request for a few minutes and use a hash key to create an opaque next link.

The first option is not a good idea, paging should be supported for all requests that deliver features.

The third option also will not work, if the deployment consists of multiple instances, the next-queries would need to be stored in a shared store.

The second option is not ideal either, but probably the best approach and could be useful in other cases, too. The approach should also be discussed in OGC (maybe just remove the links for now and wait for that discussion). There is also a somewhat related I-D.

[cportele]

PR #958 addresses the bug (incorrect links) by removing the links and documents the limitation.

How to handle paging links in responses to ad-hoc queries should be first discussed in OGC. The issue will be placed in the backlog until then.