| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
damned mailer! now I got it please apologize -------- Original-Nachricht --------
Am 10.06.2011 14:40, schrieb J. Wolff von Gudenberg:
> John
> here are some comments on your position paper
> 1. The sentence
> Note that a decoration primarily describes a property, not of the
> interval it is attached to, but
> of the function defined by a section of code that produced that interval.
> is very important and helpful for understanding. It should be emphazied
>
> 2. The following sentence is also important for your position, but I
> dislike it:
> In the remaining case
> c' = bnd, we require instead c'' = min dec(f; y) over all bounded y
> \subseteq x, to match the semantics of
> the bnd decoration.
> - as a user of p1788, I do not understand it
> - as an implementor of p1788 I see no rules or formulas how to
> procede
> - as a mathematicien i miss the proof
> - as an intervaller I think, we live better without it and always
> propagate the min of c' and c''
>
> 3. There are some typos around formula (14)
>
> 4.In the list (ii) d is the input shouldn't it be e
>
> 5-Table 5 for division is either false or I can't read it
>
> All in all I fear that p1788 will fail, if we will continue to define
> it by the theorem proof style.
> Like 754 we should define our datatype : Decorated Floating point
> interval and its operations given by detailed tables or rules. Then we
> can and we should present our assertions and proofs.
>
> Marco and me will provide some material very soon
>
> Jürgen
>
> P.S:
> John
> I do like decorated intervals and appreciate your work very much,
> but I am afraid that it will lead into the wrong direction.
>
--
- Prof. Dr. Juergen Wolff von Gudenberg
o Lehrstuhl fuer Informatik II
/ \ Universitaet Wuerzburg, Am Hubland, D-97074 Wuerzburg
InfoII o Tel.: +49 931 / 31 86602 Fax ../31 86603
/ \ Uni E-Mail:wolff@xxxxxxxxxxxxxxxxxxxxxxxxxxx
o o Wuerzburg
|