On Sun, Apr 15, 2012 at 1:07 PM, Binyamin Dissen <[email protected]
> wrote:

> On Sun, 15 Apr 2012 11:23:51 -0400 Henry Schaffer <[email protected]>
> wrote:
>
> :>On Sun, Apr 15, 2012 at 3:19 AM, Binyamin Dissen <
> [email protected]
> :>> wrote:
> .
> :>> To me it is quite logical that the verification should involve the
> verifier
> :>> writing their own program. If the same program is used then the same
> :>> results
> :>> would be expected.
>
> :>  Logical, but is it feasible? I doubt that I would review a paper for a
> :>journal if it meant having to spend a year, or even a month,
> :>writing/debugging/testing my own version.
>
> If the person cannot write the program from the algorithm, how can he
> possibly
> review someone else's implementation of it?
>

  What is a reviewer's responsibility? Quite some time ago, I entered into
some correspondence with the journal editor who sent me a paper to review.
(This was in an experimental area of science.) The editor was an excellent
and experienced scientist - and he was emphatic that it was not the
reviewer's responsibility to guarantee the correctness of the manuscript -
nor was it even always possible. For example, several assumptions may be
explicitly made - and it may be impossible to determine if they are
correct. He went on to describe the reasonable effort which should be made
to look at the design of the experiments, the reasonableness of the results
and the interpretation, the clarity of the communication, etc.

  I understand that in theoretical and mathematical areas it may be
reasonable to verify a proof.

  The case being discussed seems, to me, to fall in-between my two
examples.  For large codes, it seems to me to be closer to the experimental
example. I can imagine the reviewer asking for a copy of the code, and
running it with the reviewer's own input. But even that may go beyond what
a reviewer can reasonably be expected to do.

  What are the choices?
1) Refuse to publish any papers which involve a large code?
2) Accept such papers with no review at all?
3) Ask reviewers for a reasonable effort in determining accept/reject.

>
> :>> Obviously the formulas used should be disclosed for peer review.
>
> :>  Yes.


--henry schaffer

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to