Re: Statistics for download ?

Caio Tiago Oliveira pisze:

> For us, here in Brazil, Claudio wants to take much
seriously the 
> statistics.
> 
> The three mirrors listed on our download page (each
download is choose 
> randomly) give us detailed statistics about the total
downloaded data.
> 
> So, the people here can use a download manager to split
the file in 
> three parts (each one with a server) and we will figure
out that there 
> was one download.
> 
> If someone clicks twice on a bouncer link to got two
different mirrors 
> to split the file, it counts as two downloads. So, it's
a lot less 
> accurate than taking into account the total downloaded
data.

The important thing, if you're doing statistics, is to find
the trend, 
not the individual figures. So anyway in the bouncer data we
can see 
some trends, and this is what we actually need. (You cannot
account for 
all possible mirrors and downloads anyway, as there are many
unofficial 
sites for many communities etc.).

I'm not saying being exact is not important, but - as
Aristotle famously 
remarked, let us be as exact as the subject matter requires

Putting 
to much work in gathering singular download data makes a
little less 
sense than finding some volunteer to write a script to draw
nice charts 
online ;)

Best regards,
Marcin

------------------------------------------------------------
---------
To unsubscribe, e-mail: dev-unsubscribeqa.openoffice.org
For additional commands, e-mail: dev-helpqa.openoffice.org

Marcin Miłkowski, 15-10-2007 20:22:
> Caio Tiago Oliveira pisze:
> 
>> For us, here in Brazil, Claudio wants to take much
seriously the 
>> statistics.
>>
>> The three mirrors listed on our download page (each
download is choose 
>> randomly) give us detailed statistics about the
total downloaded data.
>>
>> So, the people here can use a download manager to
split the file in 
>> three parts (each one with a server) and we will
figure out that there 
>> was one download.
>>
>> If someone clicks twice on a bouncer link to got
two different mirrors 
>> to split the file, it counts as two downloads. So,
it's a lot less 
>> accurate than taking into account the total
downloaded data.
> 
> The important thing, if you're doing statistics, is to
find the trend, 
> not the individual figures. So anyway in the bouncer
data we can see 
> some trends, and this is what we actually need. (You
cannot account for 
> all possible mirrors and downloads anyway, as there are
many unofficial 
> sites for many communities etc.).
> 
> I'm not saying being exact is not important, but - as
Aristotle famously 
> remarked, let us be as exact as the subject matter
requires  Putting 
> to much work in gathering singular download data makes
a little less 
> sense than finding some volunteer to write a script to
draw nice charts 
> online ;)

There are other advantages with the mirrors here, since
using bouncer 
would direct us to mirrors where we could got at most 150K/s
or 200K/s, 
when I got up to 8M/s[1] on my College from two of the three
mirrors 
that we use here (and 1Mbps from the other).

Taking geolocation into account is a desired feature on
Bouncer.


[1] - My College got a good link 


------------------------------------------------------------
---------
To unsubscribe, e-mail: dev-unsubscribeqa.openoffice.org
For additional commands, e-mail: dev-helpqa.openoffice.org