Well the same critic applies to "last here" so I'll rather stick with its criteria.
I'd probably start by storing the last n ( =10 ?) "last here" on a daily resolution.
Calculating with the above formula is easy ( diff(oldest, newest) / n ), and you just need to replace the oldest record on a new day.
| [reply] |
the same critic applies to "last here"
No, it does not. "Last here" is the timestamp of your last page load. It updates on every single request.
I just opened the SOPW page, then clicked on the notification that I had a reply, then hit "talk" in cb to clear the location, then hit "reply" to type this response. That's four "last here" updates already. After I finish typing this, I'll preview my reply (updating "last here"), maybe make and preview a few revisions (another "last here" update each time), and finally submit it (yet another "last here" update).
Depending on how many revisions I make, I could potentially hit 10 "last here" updates just on making this post.
By any reasonable definition, however, that will all fall within a single "visit".
Because of this, your suggested algorithm (at least with n=10) provides little-to-no additional information beyond the existing "last here". But it can't be fixed simply by increasing n because some days my "last here" updates once (look at SOPW, don't see anything interesting, leave) and other days it updates several dozen times (multiple replies with several revisions each, delving into older posts, doing a lot of voting, etc.). Knowing "this user's last n page loads took place over the span of m days" doesn't really tell you anything.
| [reply]
[d/l]
[select] |
You missed the "on a daily resolution" (and " on a new day") part in my reply.
| [reply] |
:) also, "last visited" is optional , it doesn't always get updated
| [reply] |
