[typo] Logs / Statistics

William A. Carrel william.a at carrel.org
Mon Aug 22 01:26:37 EDT 2005


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Aug 21, 2005, at 9:43 PM, Kevin Ballard wrote:

> On Aug 21, 2005, at 11:22 PM, William A. Carrel wrote:
>
>> I imagine you'd achieve this by fitting some magic javascript or  
>> image into the article in the RSS feed.  Really this would track  
>> the same way as article views and probably could be implemented  
>> fairly easily (and non-invasively) using the filter framework that  
>> Scott Laird has been working on.
>>
>
> That wouldn't work. That would track each time the user views that  
> article in their newsreader (assuming it supports JS), not each  
> time the feed is fetched.

Who's to say that seeing who views the article in their newsreader  
and how often isn't a statistic worth tracking?  (Bob is right about  
JS in the newsreaders, so we'll presume it's a 1x1 image bug, which  
may also not be terribly reliable.)  We're going to wind up  
encountering that same kind of trouble with the normal pages too.   
Robot visitors that don't get images or javascript aren't necessarily  
going to be tracked.  Whether or not they should is an opinion that  
differs depending on what you're trying to get out of the  
statistics.  (i.e. Are you measuring server load or human viewership.)

> What I'm really most interested in, though, is not each individual  
> access of the feed (although it would be good to know that for  
> server load purposes), but rather who's reading my feed, which  
> really means tracking each time a unique IP hits my feed.

As you said, this will require reading the server logs.  A lot of the  
people hitting your feed are going to be feed aggregators (and  
various Google spider machines looking for changes to the site it  
should pick up).  Any sort of JS/Image/etc. sleight of hand tricks  
aren't going to show anything for those.

We could probably have a well-reasoned debate about whether or not  
either the "web server served a resource" or the "some browser/reader  
just pulled the magic image" stats are going to be of more value.  In  
the end, it depends on the user.  As a result, gathering both would  
probably be advisable if we're going to go to the trouble of  
integrating it, people will end up demanding which ever one isn't  
there as a feature anyway.

In the end, even with both, there will still be some audience that  
finds built-in stats inadequate for some reason.  Maybe because it  
doesn't aggregate by IP location data, maybe because it doesn't show  
whether the page came from cache, or maybe because it doesn't  
aggregate their werewolf blog views by moon phase.  These sorts of  
complications are why perhaps this would be best left as a separate  
item that just has a content filter that inserts magic 1x1 image bugs  
for tracking but doesn't otherwise require mucking around in the rest  
of the Typo code.

There are lies, damned lies and... ;-)

- --
wac
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Darwin)

iEYEARECAAYFAkMJYg8ACgkQyWl2oUGFZkxfmQCfegYXIktU9MbwAIHCFu2k9dgH
AHAAn3VMEY9PU9Hc71md2v0Nyo4qHn0d
=d3ma
-----END PGP SIGNATURE-----


More information about the Typo-list mailing list