There is a Firefox extension, WebDeveloper. It has one menu item, outline current element. It shows position of element. Example:

"html > body #id-479 > center > table > tbody > tr > td .main_content > form > table > tbody > tr > td > ul[0]" 

for an element in this page.

"html > body#id-479 > center[0] > table[0] > tbody [0] > tr[1] > td.main_content > form[0] > table[0] > tbody[0] > tr[0] > td[0] > ul"

Secondly, roughly the same question was asked by user2000 recently in Position on Web Page. The conclusions are:

• You need to know Javascript and the W3C DOM very well to figure this out.
• There are many problems doing this across different browsers.
• It's 100x harder if the user can click a TextElement (rather than just Type3 DOM nodes)
• It's 1000x harder if the user can make a text selection spanning multiple DOM nodes.

And that's just the content-targeting part. The remainder is communicating the "path" string to your perl back-end via AJAX (XMLHttpRequest, etc).

Update:

I see that like YourMother, I assumed you'd want to remember the XPath-like position in order to translate it to a (browser-relative) pixel-position.

If that's not the case, then you can start with something like YourMother has written below (using jquery, or some other library that smooths over browser differences).

To be accurate, you will need to form an XPath-like expression, so why not use real XPath?

There's a cross-platform library available here.

To construct the XPath expression, you'll have to walk up the DOM from the target element to the document.body (root). At each level, you'll need to determine 'id', 'class' and sibling-rank (with that selection priority).

-David