Try this first; use activity:subject->atom:title only as a fallback.
The code that output activity:subject was removed 2013-10-08,
and it it was deprecated for years before that....
This really should be a UUID or something else totally unexpected
but I figure that crc32 is good enough for now. The reason we keep
the main structure is because some third party scripts have begun
relying upon the tag URI format to parse out domain name, type etc.
This means we import the URI string from remote instances to track their
conversations and are able to stitch together replies in a single thread.
We might have to try to avoid collisions so noone remotely can predict
conversation URIs which we generate on our server, causing a DoS kind of
problem.
If we know the URI sent from the remote party, and we don't know the
notice it is replying to, we might still be able to put it in the same
conversation thread!
Mainly because the atom:link element requires a "web resource" but we
wish to supply a URI which might not be HTTP. We'll leave the old
atom:link element however since it's in the OStatus 1.0 Draft2 docs
and nothing newer has been released yet.
At the same time we change this to call ActivityUtils::checkAuthorship
instead to let the retrieval/verification go through event handling.
rozzin (Joshua Judson Rosen) found this error. Thanks.
We need to look up a feed profile for HandleFeedEntryWithProfile events,
regardless of whether they're an OStatus user, group, or something else;
this is the least hairy way of doing that--the alternative being
to keep spreading the same logic all over the calling code.
Theoretically, this change might allow OStatusGroups to be recorded
as the authors of activities if they pass through any authorless
activities; but that's why we have checkAuthorship().
Similarly to what ActivityUtils::checkAuthorship does; try to ensure
that activities from ambiguous OStatus feeds (groups and peopletags)
that require explicit authorship don't get in without explicit authors.
I think the migration from core to plugin is done now for DMs.
This is required since we support the Twitter-based API by default,
which is implemented in many of the mobile clients etc. But you can
disable the DirectMessage for your instance of you wish, of course.
The last commit found a bug that a previous move of the FavCommand
into a plugin called $this in FavoritePlugin instead of the instance
of a CommandInterpreter class. Now that call is statically defined in
CommandInterpreter and appropriately called in FavoritePlugin.
php5-gmp is an added requirement from StatusNet 1.1.1,
our BigInteger library doesn't seem to work well without it
and bcmath was unavailable at least in Debian testing repo.
