[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] using problems for regression tests
From: |
Gunnar Farneback |
Subject: |
Re: [gnugo-devel] using problems for regression tests |
Date: |
Fri, 28 Mar 2003 18:03:23 +0100 |
User-agent: |
EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/20.7 (sparc-sun-solaris2.7) (with unibyte mode) |
Evan wrote:
> I was wondering what our policy is on using problems from books for
> regression testing.
To begin with there is absolutely no problem *using* problems from
books for testing the engine. The problem comes with *distributing*
them. Of course the latter is usually critical for our purposes but
it's important to be aware of the distinction.
> I notice we have some problems from Get Strong at the
> Endgame; would finding problems Gnu Go has trouble with and entering them
> in be considered generally acceptable, or are things more complicated than
> that?
Things are a lot more complicated than that. Without being an expert
on copyright law (which to some extent also varies between countries),
I consider the following cases safe enough:
* Problems taken directly from sufficiently old sources are okay.
Exactly what sufficiently old means is somewhat difficult to say,
however. Currently Swedish law says that copyright protection ends
when the author has been dead for 70 years. I believe that the
situation is similar also in other parts of the world, including a
disturbing trend that the length of protection is extended (it used
to be 50 years in Sweden) any time "Mein Kampf" or Mickey Mouse gets
close to slipping into the public domain.
* The *ideas* in general and common problems can be used. For
example most of the test cases in ld_owl.tst are common life and
death problems which occur in lots of life and death books as well
as in real games. Compared to the books the configuration of
surrounding stones is different (no go book author would ever
consider using such an artificial position). However, there is still
the matter of potential copyright on the selection of problems,
etc., so one shouldn't organize the problems in the same way.
The commenting text in books is of course off limits but not
interesting for our purposes anyway.
* Problems from books where the author (or more generally copyright
owner) has given explicit permission.
* Unusual or complex problems had better be avoided. However, there
are some fair use provisions that may be invoked to include sample
problems verbatim from other sources. Still we shouldn't rely on
this too much.
/Gunnar