There are only types 'search', 'blacklist', 'referer', 'direct' and 'spam' now for hits.

'rss' is an agent_type, like 'robot', 'browser' and 'unknown'.

To tag an hit as from user-agent 'rss' there was a wrong check that does just check for 'rss', 'rdf' or 'atom' in the $ReqPath - with Phoenix it moved to $ReqURI (as $tempskin param).

I've now fixed it by checking the $tempskin param.

Actually, with Phoenix and before all hits on a blog with 'rdf', 'rss' or 'atom' somewhere in the ReqPath must have been logged as type=rss...!

if its any help i was having mass atom hits from the user agent:

Mozilla/4.0 (compatible; Google Desktop)

But wouldnt making it so a url with atom in it count as a rss feed be bad...
Like say if my url is http://allaboutatom.com

Also i just got the latest CVS build 5minutes ago, and it still has not been fixed.

I added the RSS2.0 and Atom feeds to firefox 1.5 and here is the list of the dirrect accesses:

2006-01-14 11:28:52 Del Blog All /site/b2evolutionCVS/blogs/index.php?blog=1&tempskin=_rss2
2006-01-14 11:28:20 Del Blog All /site/b2evolutionCVS/blogs/index.php?blog=1&tempskin=_atom
2006-01-14 11:26:08 Del Blog All /site/b2evolutionCVS/blogs/index.php?blog=1&tempskin=_atom

User Agents:

Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8) Gecko/20051111 Firefox/1.5 4 100.0 %

Syndication:

Total RSS hits: 0

The problem is:

A session has only one user_agent_ID attached.

A user_agent_ID is the combination of user_agent_signature and user_agent_type, so a user agent like "Mozilla/4.0 (compatible; Google Desktop)" can appear twice in the evo_useragents table: once as 'rss' and 'browser'!

But because a session is just attached to one user_agent_ID the last access for this session determines what user_agent_ID gets chosen! (Every hit has a session_ID attached)

Scenario: a user accesses the blog with a browser, the agent_type is 'browser', this gets attached to his session (which gets linked in evo_hitlog). After that he accesses the feed with the same browser. Another user_agent_ID gets used (because type is now 'rss'). This gets updated for his session and therefor the previous hit also becomes type 'rss', because in evo_hitlog just the session is linked.

The solution IMHO would be to link user_agent_ID to T_hitlog, not T_sessions: a user agent is related to the hit, not the session..

So what i understand from your post is that syndications are also decided from the user agents....

Shouldnt it work by saying ok they are doing a rss feed, so lets log it as a rss hit.
And when it isnt a rss hit it hits a user hit.

I dont see how user agents tie into rss....
As a rss feed is still a rss feed regardless of the user agent.....

Edit;
So this problem hasnt actually been fixed....
Because my blog is still getting worked by rss feeds, like 5 hits with a 2 seconds gap, then 1-5mintues wait then another 5hits with 2 seconds gap....
Its really enoying to trying to find out wether i'm actually getting any hits from people or just rss readers.

I may have been unclear, of course:

Type 'rss' is currently detected by the tempskin param (I've just committed to use $skin also instead, because it's cleaner and gets used in /xmlsrv/atom.php for example).

The problem is, that the user_agent_ID (consisting of agnt_signature and agnt_type) is linked to the session and not the hit! Therefor only the last used agnt_type in a session gets used.

For example: I go to your blog with my browser: I get a new session, with agnt_signature = 'Firefox ...' and agnt_type = 'browser'. In evo_hitlog this gets used in the hit_agnt_ID field.
Then I click on the RSS link (or use a RSS firefox plugin): a new entry in evo_useragents gets created (same signature 'Firefox ...', but type 'rss'). The session gets updated/linked to this new user_agent_ID and therefor the previous hit gets also recognized as 'rss', though it was really of type 'browser'..

I've added it to the todo list and suggested a fix, but we'll have to wait what François says, because this is no trivial change and as far as I remember he said to link T_useragents to T_sessions and not T_hitlog, when I refactored the hitlog handling.

EDIT: just for info: with KDE, my rss reader (Akregator) re-uses the session cookie, that also Konqueror uses.

Here is a screenshot of my current stats:
http://blog.balupton.nghosting.org/media/blogs/main/BlogWierdness.gif

Also if i add syndication to a feed via firefox (click the rss icon in the address bar and go add) then that also counts as a dirrect access.

Alrite it apears to be fixed in the 'Saturday, 4 February 2006' CVS build, with the only direct accesses to RSS and Atom Feeds being actuall user hits (i've tested this).

Anyone else want to confirm?

I cannot confirm, because the code has not changed in that regard.

Please refer to my last post in this thread about the problem and ask if something might be unclear.

I'm actually waiting for François' opinion on how to address this.

Ok, the people using the atom feed must of unsyndicated themselves :):(

Couldnt what you do is make it so the when a tempskin is loaded it counts a RSS Hit, and the session does not start...

So the session only starts when the user actually goes to a proper page.
And the tempskin (or feeds), don't do anything with the sessions?

Or am i totally out of sync here...

Ok the person is still syndicating my blog i was viewing the stats for the wrong one, so the problem is still there.