Re: Meta and ActivityPub

 Johannes wrote:
>
> Define an unambiguous, no-loopholes “minimum profile” of the entire
> protocol stack that guarantees a (minimum, but reliable) level of interop

Doing this would require better segmentation, probably into different
documents, of what is specified today in the three key docs.
(ActivityStream, Vocabulary, and ActivityPub).

The biggest problem today is the ActivityPub spec, since it mixes together
the specification of generic "any application" activities with those that
are more specific to a Twitter- or Mastodon-wannabee application.  For
example, a "minimal" ActivityPub specification would deal only with generic
publishing functions (i.e. To:, From: and CRUD operations - Create, Read,
Update, Delete). However, such a spec wouldn't make assumptions such as the
need or utility of an "inbox collection" -- which only makes sense in
applications that have a need for buffering of messages. In any case, even
if a real time application might have an Inbox, there is no reason why an
interoperation specification would say *anything* about the inbox unless it
was expected that entities other than the one addressed would read that
Inbox. That assumption does not hold for many applications. Similar
problems exist with the ActivityPub specification of object types or
activities (like, dislike, follow, question, etc.) that are only relevant
for a specific class of potential Activity* applications.

I would break ActivityPub into a new stripped-down "ActivityPub" that only
included the basic operations expected to be useful to the vast majority of
network applications, and a "Social Activity" spec (with vocabulary) that
was more specific to the specific style of application that was clearly the
intent of the existing docs.

bob wyman