On 23/12/2009 00:00, Kurt Roeckx wrote:
On Tue, Dec 22, 2009 at 11:29:24PM +0300, Matthieu Patou wrote:
I'm sorry to put it through David L. Mills was wrong on this.
The patch never used a TCP socket but a unix socket. And since the
initial patch the code has been reworked to be wrapped in a test
that avoid trying to open the unix socket unless the requester has
been allowed to do so in the configuration.
I guess I'm missing something here. I'm just going to make a
guess of how this works (if enabled):
1) We wait to get a UDP message
2) We send something to samba over the unix domain socket
He thought this was over TCP, and that's beside the point,
we need to wait until there is room in the kernel buffer.
This is only if there is a request for a signed packet.
3) We wait for a reply from samba, we do not process any other
packages while waiting.
4) We send a message back
5) go to 1)
And the restrict lines controls if we're going to send something
to samba or not.
Right
An other alternative could be that we don't wait for packet back but
that samba actually sends back the reply, doesn't make much
difference.
We have to wait for ntp to use event pooling as it will greatly ease the
developpement of an async behavior, that is a todo of next version as
far as I understand.
Anyway, in either case I don't see a problem with giving the admin
the option of enabling it in the config file.
My point of view is we make it available by compiling it and leave up to
the admin to activate it or not.
We can in order to avoid problem not put this option in the ntp.conf so
that only admins really willing to activate it will do (hopefully).
Matthieu.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
