[Ardour-Users] What is to be watermarked?

[Ross Johnson]

Paul Davis wrote:
> On Wed, Aug 11, 2010 at 4:06 AM, John Rigg <au at jrigg.co.uk> wrote:
>   
>> On Wed, Aug 11, 2010 at 12:02:35PM +1000, Ross Johnson wrote:
>>     
>>> Kevin Cosgrove wrote:
>>>       
>>>> Not sure the sample size of volunteered information would be
>>>> significant enough.
>>>>
>>>>         
>>> The sample size is nowhere near as important as getting a random sample,
>>> which is the biggest problem I see with these schemes.
>>>       
>
> its only an issue for certain uses of whatever information resulted.
> as a way to guage the likely size of the user community, and its split
> between linux and OS X, its probably not that important for it be
> highly random.
If you get N responses from MacOS users and M from Linux then you can 
say that there are at least N MacOS users and at least M Linux users. 
That's about all you can say and if that's what you want then ok. The 
numbers might be high enough on their own to argue for vendor support, 
or whatever the purpose is, but if you wanted to know where to put most 
development effort then it's very tenuous.

If you want to extract more from the numbers then I think that inviting 
users to register at a web page would get you a more random cross 
section (but possibly smaller sample size) because the response is 
independent of the system they use or the way they acquired the product. 
You may be able to do both ideas (phone-home and registration) to get a 
higher confidence level.

> to distinguish between linux distros, it might be more
> important for it be genuinely random.
>   
And I suspect it's not that important to be able to do that either.