Re: [gNewSense-users] Copyright question about net/irda/irqueue.c

[Sam Geeraerts]

Bake Timmons wrote:
> What do you think of this excerpt of a license comment for
> net/irda/irqueue.c ? 
>
>  *     This copyright does however _not_ include the ELF hash() function
>  *     which I currently don't know which licence or copyright it
>  *     has. Please inform me if you know.
>
> The function hash(), which I have included below, does not amount to
> much, but I do *not* think it is trivial enough to escape needing a
> copyright.  What do you think?
>
> /*
>  * Function hash (name)
>  *
>  *    This function hash the input string 'name' using the ELF hash
>  *    function for strings.
>  */
> static __u32 hash( const char* name)
> {
>         __u32 h = 0;
>         __u32 g;
>
>         while(*name) {
>                 h = (h<<4) + *name++;
>                 if ((g = (h & 0xf0000000)))
>                         h ^=g>>24;
>                 h &=~g;
>         }
>         return h;
> }

I have found a thread about it on the debian-legal mailing list [1], 
which suggests using the algorithm implementation from glibc. I guess 
that didn't happen, because I downloaded glibc from GNU and found that 
the code in sysdeps/generic/dl-hash.h is similar, but not the same. That 
file has a comment:

"This is the hashing function specified by the ELF ABI."

So the algorithm is probably well known and kind of standardized. I also 
found some other similar implementations on [2] and [3], one of which is 
licensed as GPL. I don't know if the above hash function can be 
considered as a trivial implementation, but it certainly is trivial to 
adjust [3] so that it becomes this function and it predates irqueue.c.

It's probably best to contact the FSF about what is trivial and what 
isn't. If we have to remove it, there's plenty of alternatives to choose 
from.

[1] http://lists.debian.org/debian-legal/2002/11/msg00248.html
[2] http://www.scalabium.com/faq/dct0136.htm
[3] http://www.brics.dk/~btools/ELI/libelf-0.6.4/lib/hash.c