Re: [CAD_CAM_EDM_DRO] Re: Optimizing cutting paths - The proof

In article <3B917FA6.6F259168@...>, Jon Elson <elson@pico-
systems.com> writes

>I probably shouldn't pick this nit, but the protocols do not guarantee
>correct delivery to an absolute degree. They do guarantee that all
>data that passes the CRC tests will be accurate to one part in 65,536
>because the checksum is only 16-bits. Actually, the error could be
>even more likely, if the error had a particular pattern to it related to
>the simple way the checksum is calculated. But, it seems pretty reliable,
>anyway.
>

Naturally, there are no real guarantees. The TCP/IP checksum is quite
weak but the lower level protocols are likely to use relatively good CRC
algorithms that will detect all single bit errors, all odd number of bit
errors and most if not all double bit errors as well as error bursts of
length less than the length of the CRC polynomial. The probability of
errors propagating, end-to-end, undetected is very small. Their
calculation is left as an exercise for the reader.

But you are right, I should never probably have opened this up in the
first place. Sometimes, late usually, I get a bit steamed when I see
downright disinformation propagated as if gospel and accepted by so many
as truth. I am more calm today :)


Peter Harrison
http://www.cannock.ac.uk/~peteh