[COMPONENTS] Filesystem import tail loop vs. twisted reactor loop

[kozmaz87]

If this is a requirement then a concurrent solution needs to be found. Either wire the tail functionality into twisted to make the reactor schedule the tailing loop or create multiple threads to keep the filesystem implementation independent from twisted. Either can be a can of worms in terms of dealing with concurrency.

[nigelmegitt]

Probably better and simpler to create a separate filesystem tail loop thread, as long as we can schedule an asynchronous on_new_data() method on the main thread. Creating a whole event loop is not something we should be doing though - perhaps there's something simpler than twisted already out there for this?

[nigelmegitt]

Actually I have no objection to using twisted here. It's fine. Why would we bother with the complexity of integrating a whole other backend, when there's no strong requirement to, or benefit?

[nigelmegitt]

Bumping to Release 4 since this isn't required to conform to the published specs.