|
|
Posted by Greg D. Moore \(Strider\) on 10/01/71 11:25
"--CELKO--" <jcelko212@earthlink.net> wrote in message
news:1125513734.900268.311850@g49g2000cwa.googlegroups.com...
> >> you must be aware that web requests, like most other events in
computing, are not ever logged as durations, but as instants.<<
>
> Do not confuse the recording of the data with the data model.
No one is.
> Think
> about a sign-in and sign-out sheet or timeclock. Each line is "half a
> fact"; the whole fact is the duration spent on the job. In this case,
> the user logs onto a site, stays there for x-minutes. He is not there
> for a Chronon (that is the term for a point in time in temporal
> databases).
What you're missing is most website log information is stateless. The
concept of a "duration" doesn't necessarily exist with webpages.
I.e you go to www.google.com and get a page.
Google records you requested a page. They have no idea how long you look at
it. You could get up, go have lunch, go for a walk, etc.
Shut down your computer, go to a different site, etc.
Celko, I suggest you go look at the logs of a webserver sometime. A
webpage is recorded as an instant in time.
Yes, one can try to model a visitors travel through a site, but one is not
necessarily modelling reality. They may pull up a page, go away for 5
minutes, and hit a link.
From that you can derive a "duration" they were on that page, but not
necessarily.
As I said, they could close their browser. You record no duration.
They could click a link to another site, nothing gets recorded in your logs.
Again, no duration.
They could simply type in a different URL, nothing gets recorded in your
logs. Again, no duration.
Or, they could go out for lunch, and come back and hit another page on your
site. But does that really mean that they spent a duration of an hour on
your site? Not raelly.
> So his table MIGHT look like this:
>
> CREATE TABLE Browsing
> (user_id VARCHAR(30) NOT NULL,
> website VARCHAR(255) NOT NULL,
> login_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP NOT NULL,
> logout_time TIMESTAMP, -- null means still there
> CHECK (login_time <= logout_time), -- less than one minute problem
> PRIMARY KEY (user_id, website, login_time));
>
> The real problem in this situaiton is having to round to the minute.
> We can force a convention on the stopping time to keep it away from the
> starting point by a bit less than one munute -- ('yyyy-mm-ddThh:mm:00'
> to 'yyyy-mm-ddThh:mm:59.997').
>
>
> Ever read the paradoxes of Zeno? He went thru what happens when you
> believe in Chronons. He lived in a time when Greek math was "integer
> only" and without a continuum.
>
Navigation:
[Reply to this message]
|