Use of `"@vocab": "_:"` in ActivityStreams 2.0 (at least)

[BigBlueHat]

I've been digging through as many @context files as I can find "in the wild"--especially ones in wide/popular/deployed usage--in hopes of better informing our changes and "upgrade" planning.

While digging, I uncovered this use of @vocab at the top of the ActivityStreams context file:

{
  "@context": {
    "@vocab": "_:",
    "...etc...": "...etc..."
  }
}

This ultimately means that anything not defined in ActivityStreams 2.0's Vocabulary or not covered by an additional "extension" context definition will be mapped into a blank node.

There's some well informed discussion of the risks, rewards, and reasoning behind their choice:
w3c/activitystreams-testing#4
(follow links there to past issues/resolutions).

Most of the discussions center around extensibility and some of those issues relate to the current work going into @protected (and friends).

[rubensworks]

Semi-related to this, I have a use case where I'm overriding @vocab to throw an explicit error on terms that could not be expanded when transforming to RDF (to avoid silently ignoring them).
In this context @vocab is set to "unknown://".
After converting JSON-LD documents to RDF with this context, I iterate over all triples to detect terms starting with unknown://, and throw an explicit error for them.

[iherman]

This issue was discussed in a meeting.

• RESOLVED: closed #183 as wontfix

View the transcript

4.2. Use of "@vocab": "_:" in ActivityStreams 2.0 (at least) #183
Benjamin Young: #183
Gregg Kellogg: We haven’t removed support for blank nodes, just marked it as archaic and possibly source of warning
… we haven’t broken activity streams, but will issue warnings
Benjamin Young: whole point of this thing is to catch things that might show up in activity streams. If you do consume as json-ld, you do preserve the author’s intent …
Benjamin Young: w3c/activitystreams-testing#4
Gregg Kellogg: if we look at the import activity streams an set @vocab to some urn, that way it doesn’t lose anything and it round-trips, which maps to issue above…
Proposed resolution: closed #183 as wontfix (Benjamin Young)
Benjamin Young: +1
Benjamin Young: they use their own media type in 1.0 to support list of lists. Should there be a follow up issue to say what should be done instead of blank node?
Ivan Herman: +1
Ruben Taelman: +1
Harold Solbrig: +1
Tim Cole: +1
Resolution #2: closed #183 as wontfix

[BigBlueHat]

Turns out we shouldn't have closed this one...as it relates heavily to #37.

We can continue the conversion in #37, but wanted to tie these two issues together for posterity (at the very least).