I believe the most obscure statistic that hasnt been mentioned, which is desirable, especially when one has multiple feeds, is an aggregate of the amount of actual video downloads against the number of xml page views against the number of html object views, something I would have to calculate by hand when I'm in the mood (never).
The outcome equates with exactly how many people are downloading videos in an aggregator and how many people came to the web site to view the video during any given time integral.
Where,
x = total number of xml page views
y = total number of html page views that contain embedded video objects
z = total number of actual video file downloads (with a like-sub-level of complexity in 'attempted/incomplete' file downloads).
Any other ideas? My best estimates now rely on looking at the data visually for consistency over time and trying to gauge a raw percent.
related notes:
+Just because you have a subscriber, does not mean they used an aggregator to retrieve your videos today.
+Text subscribers and video subscribers are different breeds - the text subscriber likely hits xml, html and video, where a local video aggregator may only hit xml and video while some incoming links are hot right to the video, with no html or xml.
On Nov 11, 2005, at 8:00 PM, David Meade wrote:
I created a system to track / estimate subscribers on my site. I still consider it testing, but I think it works pretty good. I dont use log files however.
Rather than parse log files, I wrote a few php functions that are called each time the feed is acccessed. It then writes (or updates) a record of the request to the databse. I have another table of "rules" that the system checks these requests against to determine how it should record/count the feed request.
You could do this with the raw logs though I would think ... so long as you have a script that has read access to them. I'd assume you'd have to read each line in and break it apart with greps.
I like my setup so far because its light weight but really flexible. You'll probably want some logic other than just IPs. (I've found that sometimes IP doesnt matter at all, sometimes its the only thing that really matters.) Having some table/list of rules lets you apply pretty good logic to whats recorded and how it's counted; for example some user agents (such as bloglines) include in their agent string the number of subscribers they have for your feed. If the user-agent is a known bot, I dont care what it's IP address is - I set its ip to "Any" (so that it will match future requests by this bot) and set subscriber count = 0.
I record all sorts of stuff for each request, but the main fields I use for 'counting' are http-user-agent, ip-address, timestamp, subscriber count.
I'm still tweaking my system, but I think it's fairly accurate. I have a link on my site (that only I can see) that I can click to get all the details of all currently counted subscribers ... and when I go over it 9 times out of 10 it all looks pretty good. Once in a while a new bot will pop up or something ... so I just go and adjust the rules as needed to ensure it's appropriately taken care of.
... not sure if that helps or not ...
- Dave
On 11/11/05, Adam Quirk <[EMAIL PROTECTED]> wrote: Does anyone know how to approximate subscribers like Feedburner does?
I assume it counts each time a unique IP address accesses the feed.
Anyone know how to do that with a raw log file?
I use AW Stats, and there no way to determine who is accessing what from their interface.
Thanks knowledgebase,
Adam
SPONSORED LINKS
YAHOO! GROUPS LINKS
--
http://www.DavidMeade.com
YAHOO! GROUPS LINKS
SPONSORED LINKS
YAHOO! GROUPS LINKS
