this patch implements those major features:
- `publish` is implemented on virtual pubsub service, thus XMPP entities can now publish
to AP using this service
- replies to XMPP items are managed
- `inReplyTo` is filled when converting XMPP items to AP objects
- `follow` and `unfollow` (actually an `undo` activity) are implemented and mapped to
XMPP's (un)subscribe. On subscription, AP actor's `outbox` collection is converted to
XMPP and put in cache. Subscriptions are always public.
- `following` and `followers` collections are mapped to XMPP's Public Pubsub Subscription
(which should be XEP-0465, but the XEP is not yet published at the time of commit), in
both directions.
- new helper methods to check if an URL is local and to get JID from actor ID
doc will follow to explain behaviour
rel 365
- cache only latest 20 items when disco infos are hidden and thus we can't discover
feature/RSM implementation
- better handling of args when calling `storage.getItems`
rel 365
- update psSubscriptionsGet to use serialised return value
- implement XEP-0376 Pubsub Account Management
- implement XEP-0465 Public Pubsub Subscriptions
- CLI `pubsub` commands updated accordingly, and added `--public` flags to `subscribe`,
`Subscriptions` and `node Subscriptions get`
⚠ `XEP-0465` is speculative, the XEP has been accepted by council but not published yet.
As is should be the next one, and current latest one is `XEP-0464`, `XEP-0465` has been
anticipated.
rel 365
- HTTP signature is checked for incoming messages
- AP actor can now be followed using pubsub subscription. When following is accepted, the
node is cached
- replies to posts are put in cached pubsub comment nodes, with a `comments_max_depth`
option to limit the number of comment nodes for a root message (documentation will come
to explain this).
ticket 364
this is necessary to do a subscription with a component managing a virtual JID (i.e. not
the JID of the component itself, but the one linked to a account, see
`SatXMPPComponent.getVirtualClient`).
rel 364
- convert a combination of JID and optional pubsub node to AP actor handle (see
`getJIDAndNode` for details) and vice versa
- the gateway now provides a Pubsub service
- retrieve pubsub node and convert it to AP collection, AP pagination is converted to RSM
- do the opposite: convert AP collection to pubsub and handle RSM request. Due to
ActivityStream collection pagination limitations, some RSM request produce inefficient
requests, but caching should be used most of the time in the future and avoid the
problem.
- set specific name to HTTP Server
- new `local_only` setting (`True` by default) to indicate if the gateway can request or
not XMPP Pubsub nodes from other servers
- disco info now specifies important features such as Pubsub RSM, and nodes metadata
ticket 363
- `NS_ATOM` is now set at class level, so it can be accessed by other plugins
- set `author` to `None` when no `name` element is found
- make `data2entry` a corountine
- fix incorrect use of yield in `trigger.point`
if an invitation is done on an email which has already received an invitation, it is
re-used and no new XMPP account or Libervia profile is created (the name is modified
though).
This is currently done in an inefficient way due to `LazyPersistentBinaryDict` limitations
(all invitations are loaded and checked to find the email). A more efficient check should
be easier to implement with storage changes in 0.9 .
`SIZE_FULL_SCREEN` may result in large files, and BoB is currently used to transfer them,
resulting in potentially large stanza. As a result, Libervia profile can be disconnected
from server if stanza max size is reached.
As a temporary workaround, `SIZE_BIG` is now the maximum size used. This should be fixed
in 0.9 with a proper way to transfer large thumbnails (i.e. not BoB).
our own jid is added automatically by OMEMO library, but during the generation of the
stanza, our list of jids was used to get the keys (and thus our own jid and devices keys
was missing).
fix 402
`callFromThread` is used instead of `callLater` with a 0 delay, as the later is not made
to be called outside of main thread (which was resulting in huge and random delays during
calls).
fix `OmemoStorage._callMainThread` which was returning wrong Deferred due a late lambda
evaluation.
When requesting disco info on a bare jid which is not in our roster, server may return
"Service Unavailable" (to avoid leaking valid JIDs). In this case, the initiation was
failing, this is now fixed by using empty categories in this case.
When text syntaxes was initialised, `syntaxes` dict was filled, but it was a class
attribute. As a result, re-initialising it (which may happen for tests, and maybe in the
future dynamically), was not starting with an empty dict, but with an already-filled one,
resulting in conflicts.
This patch fixes it by moving `syntaxes` to an instance attribute.
