Set referer for getting referring page title to nothing, then avoid fetching empty referers

[benlk]

Changes

This pull request makes the following changes:

• When looking up the referer for a hit to the tracking pixel, don't include a referer in the lookup
• If the referer URL is empty, avoid parsing it for a URL host and URL path.

Why

For #80 and #91, though I'm not sure that this does actually address the DOS issue. It will prevent the tracker from firing a request if the referer is not set, and will prevent the tracker from tracking hits without a referer.

Testing/Questions

Features that this PR affects:

• the tracker pixel's referer lookup

Questions that need to be answered before merging:

• Is this PR targeting the correct branch in this repository?
• Does it work?

Steps to test this PR:

0. Set up a Republication site, including Google Analytics
1. View your site's access log, or
2. Copy the republication text, including image, and put that HTML in a file on a server somewhere. It can be a localhost page or a .test development server or your-private-site.example.org or a GitHub page or a live-edited browser. The HTML from the republication text needs to be rendered so that the image is fetched.
3. View that HTML in a browser.
4. Open the republished article's edit view, and view the number of hits it received. It should have received one, from the URL that you hosted it. Same with your Google Analytics.
5. Extract the tracking pixel img's src="" URL, and load that URL directly in your browser.
6. You should not see any new hits for an entry in the hits table for a row with no source URL. Same with your Google Analytics.

Additional information

Requested by the Center for Public Integrity.

[jonathanstegall]

Is it best to wait for this to get merged before using this plugin? I'm finally getting around to hopefully launching it soon.

[benlk]

Probably yes. You could maintain a private fork of the plugin that has this branch merged.

[adekbadek]

Hi @benlk – could you please resolve the conflicts here?

[benlk]

@adekbadek Resolved, but please re-review this PR as it's been a long while since I worked on this code in earnest.

[benlk]

Note that this has significant merge conflicts now.