[ntp:questions] Should ntpd log failure to syslog?

Maarten Wiltink maarten at kittensandcats.net
Fri Dec 12 08:35:40 UTC 2008


"Richard B. Gilbert" <rgilbert88 at comcast.net> wrote in message
news:7qqdnST3fqThWNzUnZ2dnUVZ_sjinZ2d at giganews.com...
> Uwe Klein wrote:
[...]
>> Doing a FAIL save setup is orthogonal to announcing failures (early).
[...]
> Sorry, it's the "orthogonal" part that's bothering me.  My dictionary
> says "pertaining to or composed of right angles".  It's frequently used
> as a buzz word but seems to be without content in the context of NTP.

It's related to the idea of a 'base' set of vectors. In 2-dimensional
space, you can construct any point as the scaled sum of two vectors.
The usual choice is (1,0) and (0,1): unit-length vectors along the
two axes. The undecorated reality is that this is the definition of
the axes by the base vectors rather than the other way around.

A point (X,Y) can then be constructed from the base vectors as
X*(1,0)+Y*(0,1). Note the difference between scalar and vector
arithmetic, and the duality between base vectors and coordinates.

But you could still construct any point if the base vectors were (1,0)
and (1,1). However, since those base vectors are not _orthogonal_,
moving a hypothetical point along one of them changes the point where
it projects onto the other base vector[0]. For orthogonal base vectors,
motion along any of them leaves 'the other coordinate(s)' unchanged.

Obviously this is not rigorous but I hope it catches the idea.

The idea is: 'orthogonal' means 'unrelated' or 'independent'.

Groetjes,
Maarten Wiltink

[0] Or the plane described by the other two vectors in 3-d, and
    similar for higher-dimensional spaces.




More information about the questions mailing list