Hi Marian, Marian Hettwer wrote on Wed, Oct 04, 2006 at 11:08:11AM +0200:
> Ingo Schwarze wrote: >> I doubt the project is worth the effort at all. >> Whatever numbers might result will be heavily biased. To clarify: As far as i understood, BSDstats intends to measure the number of *BSD systems in production and connected to the internet. Should the ratio between existing and counted systems be different for different flavours, i would call that a bias. Thus, the main point of my comment was this: Due to cultural differences in the different *BSD communities, i seriously doubt that BSDstats can reach its stated goals, in particular while following the demonstrated design principle. > Of course it's biased. It's statistics of running *BSD systems. In fact, it rather is a statistics about people running the BSDstats software. So it's more a statistics about BSDstats itself than about operating systems. > How could that possibly not biased?! Well, a measurement of OS usage could be done with easily controllable biases by classical offline surveys; or you could try to use OS fingerprinting on large proxy servers (though controlling biases will be more difficult that way); ororor... Maybe such studies already exist - i did not check that. >> BSDstats is typical bloatware that lots of OpenBSD >> users will hate (not all, mind you, but many more than >> e.g. in Linuxland). > Why could bsdstats be bloatware? [...] Well, bloatware in the *absolute* sense of "software that does not add to the functionality or productivity of the server, but that does add to its complexity and maintenace effort". Some *relative* sense of "software that is more complicated than it needs to be to fulfil its purpose" was not intended. I explicity do not comment on BSDstats code quality. >> Besides, the OpenBSD community tends to just not care >> about marketing. OpenBSD is about correctness, simplicity, >> freedom and security. Popularity is *not* among the > But OpenBSD needs funding too. And popularity is one instrument > of a whole lot to get funds. Keep that in mind. Clearly a valid comment. But I doubt that BSDstats will be very helpful in that respect, but that's only a personal guess. To impress people, you would in particular need to be able to compare *BSD usage numbers to usage numbers of other OSs. As far as i see, BSDstats does not even try to do that. >> Thus, i should expect the following attitude from typical >> OpenBSD users: A software for measuring popularity? How > It's your attitude! Admitted. >> boring. What, it will even run cron scripts and open >> network connections? No way on my machine... > uuuhhh... Security Issues, hm? Well no, rather KISS (keep it stupidly simple). Don't run things unless they are needed, in particular when it comes to things like cron and daemons, yet more network daemons. On a typical server, i have a handful of networks daemons (e.g. ftp-proxy, ident, ntp, sshd on one or nfs, imap, ntp, sshd on another) so one additional is a 25% increase. I usually have exactly one cron job on a machine, so one additional is a 100% increase. Of course, KISS is *also* a matter of taste - but it is far from *only* a matter of taste. Even if it were only a matter of taste, it would be relevant as far as people's taste could bias the intended measurement. Whatever, as far as i am aware, at least during the last few weeks, nobody brought this possible flaw in the BSDstats project design to the attention of the software author(s), so i mentioned it. I have little desire to turn this into a flame or or even to discuss it much further... The BSDstats team will certainly come to their own conclusion whether the possible bias i pointed out seems dangerous to them or not.