Wikit (the software behind the wiki) is sufficient to the task, without being "Best of Breed". Wikit works well for a number of reasons, including:

• small and fast (especially when combined with Wub)
• can support web, cgi, local and embedded modes of operation
• content stored in a single database file
• openness - it tries hard to make it easy to contribute content

However, people regularly contribute ideas for enhancements that could make Wikit better without exerting significant development effort - the so called "low hanging fruit".

This is the place to document such ideas.

---

"Most recently viewed" page

At present, the only way to get an indication of what is new and interesting on the wiki is to follow the "Recent Changes" page. But this has a couple of problems - it doesn't inform as to what people are looking at, and it doesn't distinguish between new content and edits of existing content (e.g. to correct a typo).

One suggestion for addressing this is a "Most recently viewed' page that shows what people are looking at - ordered by frequency.

To implement this will require Wikit to record a view counter for each page (ignoring bots if possible), and some way to age the counter so that only views over the (say) previous week would count.

jdc I propose to add a view to wikit.tkd with the following views:

      mk::view layout $db.views {page:I date:I}
      mk::view layout $db.edits {page:I date:I who}
      mk::view layout $db.created {page:I date:I}
      mk::view layout $db.searches {key date:I}

Storing the ip address of the user may be useful to do some filtering.

The tables should be filled when viewing/editing/creating/searching. Based on these, it becomes possible to generate the most viewed reports. Some maintenance will be required. These new views will have to be truncated on a 'longest report period' period.

Is there any info on how many pages are viewed/edited on a daily basis? This would be useful info to estimate the size of the above views.

jdc I implemented the above. The views table is filled using a ::Httpd::rqCallOut callback, the edits table from the /save proc, the created table from LookupPage and the searches table from search. 5 pages are generated using this data:

1. Most viewed pages
2. Most edited pages
3. Most active editors
4. Newly created pages
5. Top searches

Using every, these pages are refreshed and their cache is cleared. Another every loop will truncate the tables.

colin Regarding 'most active pages', one could query the cache, which tends to contain the most recently used 100 pages. Much easier than duplicating the data, but only ever a snapshot.

---

jdc The number of search results is limited to 100. Someone asked if it would be possible to show more results. I added a More search results link to the search result page which shows the next 100 matches.

---

---

Posted at Jul 26/2007 01:15AM:
Colin: Thinking about the possibility of generating ref pages on-the-fly, instead of en masse. Ah, now that I think about it for more than 10s, I see that this just won't work. I have a cold :}