Problem: How do you know the number of horses in a country? Solution: You count horseshoes and divide by four. Jokes aside any statistician will tell you that when you can't proceed to a direct estimation you estimate something that is correlated. This is particularly useful if you are only interested in evolution but if you want absolute numbers this can be done through a survey telling you how many A per B with A being the studied variable and B the correlated one. So a sensible correlated variable would be number of IP adreses that are pulling updates from such or such Fedora version. A still better one would be number of updates each time there is a security hole in the kernel or glibc since all users are using them and we can assume over 90% users will update. Both are better than primary downloads since these are influenced by DVDs in Linux magazines and people that update instead of installing afresh. Problem: These are influenced by such parameters as -Dual booting (person could be using several versions of Fedora in same box) -Distro or OS hopping (should a person who spends 10% of her time in Fedora get the same ponderation than a full time user)? -IP masquerading -Local mirrors -Tor. You will need to set aside connections from TOR nodes and keep an eye about their number respective to total. You can estimate the real number of TOR using machines starting connections through the number of updates of "must-be" packages like kernel or libc: if each time there is a kernel update you get 10 kernel donwloads then it doesn't matter if all came frm the same TOR exit ppoint or if from one version to the other you get connections en enytirely different set of TOR nodes: you know there are 10 downloading machines (I assumme there are no lazy users). While it can be hazardous to try to get abolute numbers you will a good idea about _evolution_. Just make a survey every couple years in order to know evolution of the relationship between number of users and number of downloads. Could be for instance that more and more of Fedora users use local mirrors. Now it seems Fedora's downloads have been steadily going downwards and it is highly improbable that is due to a spectacular increase of number of people setting local mirrors. Fedora has a number of problems 1) It hasn't managed inspire the same kind of fanaticism Debian has. When in Debian something is broken, downright dangerous (eg signed packages came much later ) or just s..cks it is not a bug it is a feature. Not such with Fedoora.. 2) It has an uninspiring installer. 3) A slow update manager and front ends aren't particularly attractive 4) After completing installation you have a lot more _unassisted_ postinstallation configuration than in, say Suse or Mandriva. 5) Config tools are poorer than in Suse or Mandriva(BTW did you set a "spy department" with the task of keeping watch of what others were doing?) and are not task oriented. For instance in Suse if I try to configure a Windows printer the installer will automatically check and install the Samba client packages. And, just in case Fedora does that now for printers, Suse and Mandriva do things like that for about everything. 6) Fedora hasn't managed to get the same kind of hype there is around Ubuntu. Partly because of not benefitting of DEbian's propaganda network, partly because of points 2,3,4,5 and partly due to bad PR. The splitting of Fedora in three versions could be a good thing maientance wise but if goal was to increase its pouplarity I fear that is it will be like applying a plaster on a wooden leg. ----- Mail original ----- De: "Ian Malone" <ibmalone@xxxxxxxxx> À: "Development discussions related to Fedora" <devel@xxxxxxxxxxxxxxxxxxxxxxx> Envoyé: Mardi 2 Décembre 2014 13:19:02 Objet: Re: How many users does Fedora have? On 1 December 2014 at 20:55, Stephen John Smoogen <smooge@xxxxxxxxx> wrote: > > > On 1 December 2014 at 05:05, Reindl Harald <h.reindl@xxxxxxxxxxxxx> wrote: >> >> >> Am 01.12.2014 um 12:57 schrieb Pierre-Yves Chibon: >>> So we come back to the question: is any number better than no number at >>> all? >>> Even to get a trend? >> >> >> no number is in fact better than wrong numbers backed by nothing beause >> they lead in wrong conclusions - your 122/133 numbers could in reality also >> be 1000 users installed them from mirrors and your calculation is the best >> example for wrong assumptions >> > > While that is 'true', most of the world doesn't work on 'true'. Your cars > speedometer doesn't give you the accurate km/hour. [Even BMW digital has a > +/- 2 km/hour due to all the factors from tire present size to road > conditions.]. The answer is can you accurately remove enough noise to feel > confident that you are doing 100 km/hour versus 120 km/hour. The same goes > for measuring downloads. > It depends how and why you are collecting data. If, for example, you are sensitive to how often downloads occur from mirrors, then sites that use local repositories will be under-represented. And if that happens to be a use type that typically customises their package choices in some way then those choices wont turn up. There are two types of error in general, random noise and bias. Rough analogy, if your spedometer actually reported 0.85 of your real speed if may not make a very noticeable difference at 30mph, but at 70 it does (and worse if it turns out to be non-linear). -- imalone http://ibmalone.blogspot.co.uk -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
