DROPOWNERSHIP command

[atagar]

Hi Nick! As discussed a DROPOWNERSHIP command would be handy for the following use case...

00:04 <+atagar> Sorry, I'm probably not groking what you're suggesting. 
OwningControllerFd sounds to be for TAKEOWNERSHIP of a control socket 
(https://gitweb.torproject.org/torspec.git/tree/control-spec.txt#n3396). Do you 
mean 'don't open a control port, but open a control socket instead'?
00:05 <+nickm> It is more special than that.
00:05 <+nickm> you open a socketpair, and pass one end to tor when you start it
00:06 <+nickm> since the os enforces ownership, you don't need to do any other authentication
00:06 <+atagar> Ah. So stem opens a local socket before starting tor, provides 
this socket as a torrc argument, listens to the bootstrap messages, then 
disconnects?
00:07 <+atagar> That certainly sounds like it would do the trick.
00:08 <+atagar> In this case I don't want tor to terminate when I disconnect 
(which is the usual purpose of TAKEOWNERSHIP), but if I can avoid that this 
sounds perfect.
00:09 <+nickm> sounds like a DROPOWNERSHIP command would be sufficient for your 
needs then
00:09 <+nickm> open a ticket and I'll make one :)
00:10 <+nickm> or remind me next week
00:10 <+atagar> Wonderful! Thanks, will do.

[nickm]

Implementation in branch ticket28843, PR at ​https://github.com/torproject/tor/pull/587

Can anybody else write the control-spec.txt patch?

[atagar]

Thanks Nick! Pushed a prospective spec patch to my 'dropownership' branch. I wasn't quite sure what to cite for the version or if the characteristics I describe are accurate so feedback welcome.

​https://gitweb.torproject.org/user/atagar/torspec.git/commit/?h=dropownership

When merged I'll follow this feature up with a few stem integ tests to check for some of these characteristics (idempotenticy and re-establishing ownership).

[mikeperry]

I assume the test failure is either some travis issue or other transient failure? Seems unrelated.

This looks good otherwise. The spec too.

[nickm]

Great; merged both.