Re: [PWE3] Proposed new 5.4.3 for arch draft

[Stewart Bryant <stbryant at cisco.com>]

neil.2.harrison@bt.com wrote:

> Stewart Bryant wrote 03 March 2004 04:44
> <snip>
>
> > The checksum solution does not work - there is still the 1/64K chance
> > of a clash.
>
> NH=>Is there not always a possibility of a mistake here if 
  undetected errors affect the 1st nibble solution?

Yes

  Though
  I guess if this was *always* followed up by a checksum test
  (for *both* assumed IP client and assumed X (PW) client cases)
  this would at least verify/refute the case (subject to a
  further 1/64k prob of error).  If 1st nibble neither case
  then bin?

I think that it would be better to leave it to the application
to decide how important it is to trap undetected datalink and
memory error. If it is important, then check the checksum (not
that the checksum is a very strong check) otherwise don't.
The point about the nibble vs the checksum is that if we control
the nibble, the application gets to choose the certainty of
identification.


> regards, Neil
>
> _______________________________________________
> pwe3 mailing list
> pwe3@ietf.org
> https://www1.ietf.org/mailman/listinfo/pwe3

_______________________________________________
pwe3 mailing list
pwe3@ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3