First things first: we track very very little data today. I’ve posted before that we use our Application Update Service (AUS) pings to get a sense of where our main usage around the world is and to try to spot problems when they happen (it’s notable that this is a secondary usage of that system — the primary function of AUS is to enable timely updates to the softeware we release — in Firefox 3 and future versions of Firefox 2, we’ll watch instead an analogous ping that checks whether updates to add-ons are available.) Beyond that, we don’t collect much data in the product at all. [...] It’s early days, but it seems to me at least that opening up all sorts of data — from web usage to the social graph & beyond — is going to be the topic of conversation for a long time to come.
Do they seriously think these statistics will be somewhere near reliable? ??? Or do they just want to claim big figures for sites that would be overrepresented in them? 88) Like Mozilla affiliated sites? Would Firefox fanboys buy these figures?
I don’t want to hurt anyone’s feelings nor start something nasty here (which would be off-topic so don’t go there), but Firefox users should get acquainted with the fact that there exists a world outside of Firefox. And maybe, just maybe, Firefox’s market share doesn’t necessarily have to grow rapidly until it reach 100.0 per cent.
But it’s not only that Firefox users aren’t --and, let’s face it, won’t be-- a representative sample of the total Internet population. Firefox users who willingly install this tracker won’t be a representative sample of the total Firefox users population, either.
So this new way of measuring traffic would surely be even way less reliable than the mentioned alternatives. This Michael Arrington’s got to be some really enthusiastic Firefox evangelist, and ignore even the barest basics of statistics, to write that article.
About the words however, if it’s not hidden and is installed with consent it cannot truthfully be called spyware. Even if this was to be bundled with Firefox so that you can’t install the latter without the former, ir wouldn’t be spyware if it’s up-front.
(By the way the first link’s broken.)
This is going to be really off topic, but have you tried disabling the automatic check for new versions?
P.S.: I know this post looks like an Opera vs. Firefox offensive from a know Opera user, but please disregard. ;D Seriously, it’s not.
If you strip away fanboy references and the like, your post is relevant.
The thing is, what i’m reading is so vague i don’t even know what data they want to collect, they’re merely talking about, or rather, introducing (!) the discussion.
There’s nothing really substantial to talk about (for us, here), if we look at it objectively.
But your reservations are logical.
I agree that since going to firefox 3 there does seem to be a lot of data transfer to googles IP address.This seems to happen pretty frequently and is getting to be quite annoying. You will be reading some post such as this one,and all of a sudden you will hear your harddrive churning away,you look at your active connections and see data transfer to google. I have unchecked the automatically check for updates to firefox and search engines(only the add ons is checked). Think i may give Opera another try,its just a shame they don`t have something similar to no-script,but it does seem that Firefox and Google have become somewhat joined at the hip.
Yes I’m not trying to criticize Firefox itself, only saying that the data thus collected would have no known relation to real overall traffic figures. But I do think that Mozilla isn’t stupid and has to know this, and then they must know that they’d be claiming figures biased in favour of sites which are more visited by this non-representative sample.
And on the other hand Michael Arrington did exaggerate when he said that yet another traffic surveillance device, way less reliable (actually completely unreliable) than the ones already existing, is “what the Internet needs”. I don’t think he’s into a conspiration by the Mozilla Corporation, so my guess is that he got carried away by some fanboyism of his, and did a diservice to his readers.
PS: I stopped spying myself–I mean monitoring where my programs connect–some time ago. Blissful ignorance is best. ;D
Well I too have only the info in the article, but according to it the data would be the surfing habits (in an anonymous aggregated form) of the people who are using this feature. Many other programs (Windows Media Player and many more examples) have the option to gather some kind of data about usage, I’m guessing this would be the same but specifically about web traffic, I can’t see other way to interpret the article.
But then you’d at best know the surfing habits of Firefox users, but you can’t extrapolate for the total population with that, even if you knew all about Firefox users. (Besides you’d have data about the Firefox users who agree to using this, not about the total.)
As far as I know statistical polling works like this. You want to know something about a “population”, in this case the whole of Internet users worldwide. Then you select a big enough “sample” that is “representative”. This means that each and every variable that may influence what you are polling, must have the same distribution in the population and the sample: the same proportion of males and females, the same proportion from every country, from each urban and rural environment, of education level, of age, family environment, weight… you name it. (And theorically the same correlation among all those variables too, for example the same proportion of fat middle-aged single graduated women…)
Even if your sample were representative of a population (Firefox users) that’s part of a bigger population (Internet users), statistically speaking there’s nothing you can do to extrapolate the results. You’d need to poll a representative sample from that other bigger population. It may work if you knew that the variable “browser choice” has no effect on the variable polled “surfing habits” (only the same would be smaller compared to total Internet users than to Firefox users and so less reliable), but the case is that browser choice is heavily correlated to surfing habits.
However, I myself am not that interested in knowing worldwide surfing habits. ??? (And you have Google Trends and stuff.) I guess websites are interested in measuring their number of own users, and there are already ways to measure on a per site basis.
I’m not talking about extrapolating.
BTW, although there could be other good sampling methods (i don’t know), i only studied random sampling.
Randomness is an issue with internet statistics, from what i can see. Anywhere i look, i see bias.
Frequently quoted statistics are actually just logs from a specific network or something. Like to infer browser market share, or OS. It’s not reliable.
Suppose firefox data was used as you say to extrapolate. Bias would be present no doubt.