[ntp:questions] Doubt about Syncronization of GPS SPECTRACOM(2)
mayer at ntp.isc.org
Fri Mar 9 02:13:28 UTC 2007
Rick Jones wrote:
>> Windows doesn't support it until Vista (but then I'm only adding
>> transport support for IPv6 for Windows now)! And you will see this with
>> older platforms. We had to emulate getaddrinfo() for older platforms.
> Sounds like stuffing those checks into the emulation of getaddrinfo()
> and then supporting AI_ADDRINFO to the main code would be a good thing -
> that way when someone uses the ntp code as an example they will be
> seeing goodness :)
You're right about that except that the getaddrinfo() emulation code
will never support IPv6. getaddrinfo() was designed to support both IPv4
and IPv6 but if you don't have the function in your library on your
system it's reasonable to only assume IPv4. If you look carefully, it
basically translates into a gethostbyname() wich can only handle IPv4
anyway. So why bother?
On the other hand adding it into the code to force the real code to
check if possible would be useful since people have systems set up that
can support IPv6 but didn't set up the infrastructure on the machine so
there are no IPv6 interface addresses but it can retrieve an IPv6
address as a result of the lookup.
>> Are you looking at doing performance testing for NTP like you've done
>> for BIND?
> Hadn't really thought about it - didn't think that NTP had performance
> issues. I suppose if folks think it important something could be added
> to netperf4 (which I cannot recall if it uses AI_ADDRCONFIG or not :)
I know of no performance issues with NTP even when I try and bang it
hard for testing purposes.
> althought there is also the place where I'm trying to "GObjectify" things:
> rick jones
> BTW, since I'm not subscribed to questions at lists.ntp.isc.org those folks
> only see stuff in the replies :)
We have a two-way gateway between questions and cptn so everyone sees
More information about the questions