If a given page is always SSL/HTTPS or if it is always non-SSL/HTTP, you can use page-relative, root-relative, or canonical linking on that page.
If a page can be accessed using either HTTPS or HTTP, then use only page-relative or root-relative linking on that page, unless you are generating the page dynamically with a script can dynamically determine the proper protocol to specify in canonical links.
Examples of my terminology:
Page-relative links: <img src="logo.gif"> or <img src="../logo.gif"> or <img src="../images/logo.gif">
Root-relative link: <img src="/images/logo.gif">
Canonical link: <img src="https://example.com/images/logo.gif"> or <img src="http://example.com/images/logo.gif">
It is the client (e.g. a browser or search engine robot) which resolves links on your page to the canonical form it needs to send a HTTP or HTTPS request.
If a page-relative link is specified on a page, the browser strips off the part of the URL following the last slash in its address bar (i.e. it removes the "page name"), appends the specified page-relative link, and issues a request for the resulting canonical URL.
If a root-relative link is specified on a page, the browser strips off the part of the URL following the first single slash in its address bar (i.e. it removes everything after the "domain name" and the optional port number), appends the specified root-relative link, and issues a request for the resulting canonical URL.
If a canonical link is specified on a page, the browser simply uses it as specified.
Note that only the canonical-link format can or will change the protocol.
Jim
