[ntp:questions] Re: What's different about these (S)NTP servers?

Danny Mayer mayer at gis.net
Thu Oct 30 16:20:06 UTC 2003


gabriel rosenkoetter <grosen at cc3.com> wrote in message news:<mailman.46.1067473108.1751.questions at ntp.org>...
> On Wed, Oct 29, 2003 at 02:50:53PM -0800, Danny Mayer wrote:
> > I don't know what this A.B.C.160 stuff is.
> 
> It's a rather common way of hiding my globally-addressable IP
> address space from a public mailing list (one gatewayed to a Usenet
> newsgroup, no less, and so archived in a variety of places on the
> world-wide web) so as to avoid unwanted visitors. I find it hard
> to believe you've never seen its use before.
> 
> > It has no meaning.
> 
> Sure it does. A, B, and C are variables for the same numbers each
> time. The real numbers are irrelevant, and expose my internal
> networking in a way that I view to be a security threat.
> 
I see this all the time. Mostly it prevents others from helping you diagnose
problems. In this case the numbers may be irrelevant. 
> > You should A) upgrade to 4.2.0 and try again
> 
> This is a corporate environment, I (or, more importantly, my
> management) would rather not use non-vendor-supplied software
> packages if it's at all avoidable. I'm sure you must have
> encountered this attitude before; it stems from vendors pointing
> fingers first at the usage of software other than what they provided
> (or stated support with; think Oracle) if a problem is ever brought
> to them.

In this case, the vendor is just resupplying the ntp software and may
not even provide support. It's a not uncommon situation.

> 
> > B) don't use w32time on a PC since you can't guarantee what to
> > do. Install ntp 4.2.0 on the W2K machines and turn off the w32time
> > service. Things should work between like applications. I don't know
> > what kinds of protocols that w32time supports.
> 
> w32time is well-documented to support a subset of NTP, called SNTP
> (S for simple), which Microsoft "embraced and extended".
> 
> My problem is not that the NTP versions I'm using don't work with
> w32time servers, but that they don't work with w32time on one
> specific server, while they do on others, and give me no indication
> for the reason that they don't work. This inconsistent behavior,
> across a wide number of NTP versions, leads me to ask here what my
> steps for diagnosis should be.
> 
> Your preference against Microsoft software is understood, and I
> share it, but it's bigotry to suggest that NTP, ostensibly an
> implementation of an open standard, must be used everywhere to use
> that open standard. That's very poor open source advocacy, and it's
> exactly ignoring the questions I asked. Please, if you can't respond
> helpfully, don't; you're just wasting everyone's bandwidth.
>
You are wrong. I have no bias against Microsoft software, I use it all
the time. The issue is one of standards and interoperability. I don't
believe any work has been done to ensure interoperability of Microsoft's
w32time with NTP, and in fact I have a vague recollection that Dave Mills
pointed out within the last six month (but I don't remember in what forum)
the problems with Microsoft's implementation. The bias or bigotry is that
we can only know for sure that NTP interoperates with NTP. We can't be
sure about anything else. If someone has done some testing they should speak
up as we have no information about it.

I'm not ignoring the problem, I'm trying to give you some hints as to the
issues involved. We don't have enough information to work with, but you
probably can't help there since you don't have access to Microsoft's source
code. 
 
> So, what would lead ntpdate to communicate with two (SNTP) servers,
> get seemingly functional results out of both, but refuse the data
> received from one and accept the data received from another?
> 
> Or, if that's too vague a question, what can I do to get further
> information out of ntpdate? (ntpq, unfortunately, won't work with
> these servers, as the control channel is one of the pieces of the
> implementation missing in SNTP.)
> 
> I'm glad to build and entry a newer or modified version of NTP in
> the interest of diagnosis, of course. I'll go see what ntpdate
> 4.2.0 has to say about these two servers.

At the very least it's worth testing a 4.2.0 NTP and see if that
fixes your problem, but I a feeling it's a protocol issue between two
different apps.

Danny



More information about the questions mailing list