Steve, you can open sockets and send data down them natively in Gecko-based apps... and the XPCOM components are scriptable, so you don't have to write any C/C++. So, there's no need to write an command-line program to send the plaintext down the socket (only to launch it from Torpark). Let me know if you need pointers.
----- Original Message ----
From: Arrakistor <arrakistor@xxxxxxxxx>
To: Roger Dingledine <or-talk@xxxxxxxxxxxxx>
Sent: Tuesday, August 1, 2006 11:37:17 PM
Subject: Re: Tor Question
Okay, so i COULD send the plaintext signal to the control port
have success under win32. I thought you said this, but questioned my
memory once i read the code.
The reason I ask about winking nodes is that I added a button to kill
the tor process. When torpark notices that there is no longer a
tor.exe process, it restarts it. Occasionally the tor process would
instantly exit for some reason, too infrequent/inconsistent to tell
everyone to log, but frequent enough to be an annoyance. The upshot is
that if you want to flush your tor circuit, you can simply click a
button on the toolbar and it gets you a new circuit. I did this
because often a user will be surfing and circuit fails for one
or another and they have to shut down and restart the browser to get a
new circuit. Because the process is so dirty, I am worried about how
it may affect the tor network.
I still have yet to heavily research making torpark actually talk with
tor.exe instead of kicking it around. I wonder if it is possible to
pipe these plaintexts directly to the controlport with some command
Tuesday, August 1, 2006, 10:11:59 PM, you wrote:
> On Tue, Aug 01, 2006 at 10:02:57PM -0500, Arrakistor wrote:
>> I have a question. I noticed in some code that it said if tor is
>> running on a win32 system, to disable signals. So even if I sent
signals like sighup to the process, it wouldn't be listening. Is this
> Correct. Windows doesn't have signals, so there is no concept of sending
> a sighup to a Windows process.
> We fake this by letting you connect on the controlport and send the string
> "signal hup", aka "signal reload". Then Tor behaves as though you just
> sent it a unix-style signal.
>> Also, what kind of damage does it do to the tor network for a node to
>> just instantly disappear without notifying the network?
> Any circuit that's currently using that node breaks. So if a user has a
> long-term IRC or IM connection on that circuit, she gets disconnected.
> Tor will automatically recover and build new circuits, but the application
> will have to reconnect over the new circuit.
> We try to tolerate nodes that appear and disappear frequently, by avoiding
them when building circuits for destination ports that we expect will
> want high uptime: see LongLivedPorts in your manual page. After all,
> servers will have to disappear and reappear periodically.
> But if they aren't online very long, new clients won't have time to hear
> about them from the directories, so we won't be able to make good use
> of them. This is another item on our list of hard things that ought to
> be fixed. :)
> Why do you ask?