[ntp:questions] Re: an example of a negative RootDelay, please...

David L. Mills mills at udel.edu
Sun Aug 10 00:00:42 UTC 2003


Morpheus,

Fixed-point arithmetic in twos-complement is just like integer. The only 
difference is in the scaling. Your interpretation of the root delay 
field is broken. I don't know how to help you. As for understanding how 
the delay can be negative, I repeat; slog through the error anlysis and 
note that negative delays can happen with large frequency skews and 
small propagation times.

Dave

Morpheus wrote:
> Hi,
> 
> 
>>The root delay format is described on p. 51 of RFC-1305 as I'm sure you
>>have seen.
> 
> 
> yes, i have...
> 
> 
>>As in almost everything Internet, this and other NTP packet
>>variables are in twos-complement representation and big-endian network
>>byte order. That completely describes both positive and negative values
>>except for scaling. The decimal point for the root delay and root
>>dispersion is between bits 15 and 16. Once you understand how
>>twos-complement arithmetic works, the answer to your question is obvious.
> 
> 
> i know 2-complement arithmetic for integers or floating-point numbers, but
> not in fixed point arithmetic... let me take some example: -1,1 in
> RootDelay format for integer part we got  '255 255' (in byte) but for
> fractional part? something like '128 0'? (maybe it 0,2..) or something
> else? i try search something but i can't find some stuff that let me
> understand... you may help perhaps....
> 
> 
>>Negative values for roundtrip delay and by implication root delay are in
>>principle possible, but very rare.
> 
> 
> thus they not impossible, so i have to understand how they work...
> 
> bye and thanks again




More information about the questions mailing list