In general, this whole thread reminds be of the Debian PopCon package/database. (...) choose to install on their system if they want to participate.

That works only if your user base in greater than very huge. There have been MILLIONS of Debian installations, while only less than 5000 were counted by PopCon.

scares me for a variety of reasons that will either seem obvious, or paranoid depending on the readers personality.

I'd like to know those reasons. What is wrong with letting others know your OS and version of Perl? No paths, personal information or anything non-static will be sent. $^O and $] are compiled into perl (in fact, $^O is not the platform perl *runs* on, but the one it was *compiled* on) and the module's name plus version are not computed, but hardcoded information. What is your objection to sharing this non-personal, non-identifying information, and why isn't opt-out good enough for you?

1. The benefit is not in knowing absolute numbers of installation base, it's in knowing the ratio of installations to sample size. if 50% of all participating users have Foo installed, ant 20% of all participating users have Bar installed, you should be able to gain a high level of confidence that Foo is more prevelant "in the wild" then Bar

2. At no point prior to your message has any one described a way by which this functionality could be hooked into MakeMaker, with an Opt-Out, which would be obvious to people who don't know they should look for it because they just upgraded module Foo without any idea that the new version of Foo is submitting this data.

   furthermore, I'd be impressed to here any such description that sounds more intuitive and easy for users to do then to have an "Opt-In" instead.

3. In general, I'm a big believer of "Opt-In". I rarely encounter something that's "Opt-Out" which doesn't piss me off by the nature of being "Opt-Out"

1. The benefit is not in knowing absolute numbers of installation base, it's in knowing the ratio of installations to sample size. if 50% of all participating users have Foo installed, ant 20% of all participating users have Bar installed, you should be able to gain a high level of confidence that Foo is more prevelant "in the wild" then Bar

If 0.005% of all participating users have DBIx::Simple installed, it will probably not even show up in the statistics. As I said, this is only a possibility if the total number of users is more than huge.

2. At no point prior to your message has any one described a way by which this functionality could be hooked into MakeMaker, with an Opt-Out, which would be obvious to people who don't know they should look for it because they just upgraded module Foo without any idea that the new version of Foo is submitting this data.

I have said that a message would be given, that the script would wait for 10 seconds and that the way of opting out would be made very obvious.

3. In general, I'm a big believer of "Opt-In". I rarely encounter something that's "Opt-Out" which doesn't piss me off by the nature of being "Opt-Out"

*In general, I agree. However, most of the time it is because the opting is for receiving offers, news and spam. It is because they will know my mail address or my name. It is because they might gather information and keep a file of just me. It is because *personal information* is sent.

However, I do not believe that any of the information that I wish to collect compromises privacy or security in any way. The only more or less sensitive information would be the IP address. But even that information cannot be used for anything other than statistics. Of course, I will not save the IP address, as that information is not relevant to what I want to know.

Juerd # { site => 'juerd.nl', plp_site => 'plp.juerd.nl', do_not_use => 'spamtrap' }