When using the multiple label filter, Confluence generates urls for applying additional labels to the filter. These urls are generated by adding the new label to the end of the current url. This make is possible to create multiple urls that result in the same filter being applied.

For example:

http://wikis.sun.com/label/vm+global+virtual

contains a url to add "xvm" label to the current filter and the url for such a filter is:

http://wikis.sun.com/label/vm+global+virtual+xvm

If I go to that url I can remove the "vm" label from the filter by going to:

http://wikis.sun.com/label/global+virtual+xvm

From where I can add the "vm" label back to the filter via:

http://wikis.sun.com/label/global+virtual+xvm+vm

By doing this I got to the same vie via two different urls:
http://wikis.sun.com/label/global+virtual+xvm+vm
http://wikis.sun.com/label/vm+global+virtual+xvm

This usually isn't an issue for users, but when a search crawler hits the sites it always gets caught in indexing the same pages over and over just because the urls are different. This results in poor search index quality as well as unnecessary load on the server.

Proposed solution: sort labels in the URI so that any combination of labels results in only one URI.