Steps to reproduce:

1. Insert a relative link to a Confluence page via Insert > Web Link > Address: /conf43/display/ds/test
2. Save the page and you can see that the domain address is appended to the relative path e.g. http://www.atlassian.com/conf43/display/ds/test

Confluence no longer supports writing a web link as a relative URL, however you can get around this still by inserting a non-web link. If you look at the instructions on Linking to Pages, the following ways will work effectively as relative links:

1. Adding a link using autocomplete
2. Adding a link using the Insert Link dialog (either search or recently viewed)

This kind of link does not store the base URL of your Confluence instance, it refers to the page by the (space id, page id) pair. This also means that if the page name changes, then the link will still work. If you left the link text the same as the page name originally, then it should automatically update to the new page name as well.

While the methods described above works for a single page, you cannot use autocomplete or the search to find the rss action page. For instance, you'd like to make a relative link to an rss feed such as this:

/createrssfeed.action?types=page&spaces=conf_all&title=pagename&labelString=work&excludedSpaceKeys %3D&sort=modified&maxResults=30&timeSpan=30&showContent=true&confirm=Create+RSS+Feed&os_authType=basic

The methods described won't work because the link is not a Confluence page. Using the insert link dialog, you can insert this relative link as a web link but it will be saved as a absolute link instead. This is also becoming a problem when you are trying to move a staging server to a production server. You'll have to manually change the link individually to match the current production domain.