Re: [Tails-project] some stats on the website

intrigeri:
>
> sajolida wrote (24 May 2016 19:09:43 GMT) :
>> Hits per language
>> =================
>
>> for lang in en fa fr de ; do echo -n "${lang} " ; grep -E "GET
>> .+\.${lang}\.html HTTP/1\..\" 200" access.log* | wc -l ; done
>
>> en 1501323 (83.1%)
>> fa 11468 ( 0.6%)
>> fr 124823 ( 6.9%)
>> de 170007 ( 9.4%)
>
> Note that this misses hits to pages that are not translatable (whose
> extension is .html, and not .en.html), so it artificially lowers the
> number of English pages hits. (I doubt that the difference would be
> significant, but it seems easy to adjust the script to take these
> other hits into account, and you asked for feedback about the scripts
> themselves :)

Thanks for the feedback, I didn't think about non-translatable pages at
all, but on this section I wouldn't change my script. My goal was to
answer to Draghana about the impact of the Farsi translation. So I was
trying to answer the question "How many of our visitors use the Farsi
translation (if available)?" or "Which percentage of our users prefer
Farsi over English?" so removing non-translatable pages makes sense. I
didn't mean to compute the number of page view in each language (as it
would have to take into account the state of translatable pages).

>> Top 50 pages in Farsi and their hits
>> ====================================
>
>> Note that this doesn't mean that these pages are actually translated in
>> Farsi. For example, the top 2, 3, 8, 10, and 12 pages are not translated
>> into Farsi.
>
> … this made me notice that the rendering of untranslated strings on
> Farsi pages is pretty bad. I believe this is due to the fact that
> we're declaring text direction as RTL globally, for the whole page.
> Ideally ikiwiki's PO plugin would set dir="ltr" for any untranslated
> string, but I think it's a lot of work to get it right :/

Make sense. Feel free to fill a bug upstream ;)

>> Top 50 pages across all languages
>> =================================
>>
>> grep -E "GET .+\...\.html HTTP/1\..\" 200" /tmp/access.log | sed -n -re
>> 's/.* ([^ ]+)\...\.html HTTP.*/\1/p' | sort | uniq -c | sort -rn | head
>> -n 50
>
> This also misses hits to non-translatable pages (but here again,
> I doubt many of them would be in the top-50 anyway).

Here it would indeed make perfect sense to include non-translatable pages.

>> 554957 /news
>
> Interesting. I suspect this includes a lot of feed readers, but
> I guess it's not worth investigating :)
>
>> 154156 /install
>> 146827 /install/os
>> 99661 /install/win
> [...]
>> 65685 /install/win/usb/overview
>> 62222 /install/win/usb
> [...]
>> 33440 /install/debian
> [...]
>> 20882 /install/debian/usb
> [...]
>> 20428 /install/dvd
>> 20370 /install/debian/usb/overview
>> 19971 /install/linux
> [...]
>> 19305 /install/vm
>> 14709 /install/mac
>> 12914 /install/win/clone/overview
> [...]
>> 11638 /install/clone
>> 10947 /install/download
>
> The Debian/Windows/Mac/VM ratio is very interesting!

Indeed!

> For example, I expected more OSX than Debian/Ubuntu, and I would have
> wrongly guessed that the Windows/Debian ration would be much higher
> than 3.
>
> It's also interesting to see this many DVD and VM hits.

My conclusion is that our user base is not what it should be... :(

I guess that these ratios are very different from the market shares of
these operating systems in general. But I see little reason for them to
differ as Tails runs independently from the base OS. Maybe Linux user
are more conscious about privacy than Windows user? But would we say
this about Mac? So for me, seeing this would rather reveals that our
proposal and adoption and installation processes, are off-pulling for
Windows and Mac users, maybe by reputation (eg. hardware support for
Mac, complicated to install and start, etc.).