Wikipedia:Tools/Navigation popups/About fixing redirects

Some users of Popups are making inappropriate use of the popupFixRedirs feature. They think their use of the feature is reducing work for the Wikipedia servers, but in fact, they're increasing the work the Wikipedia servers have to do.

Say you come across a perfectly legitimate redirect link, for instance (morphosyntax redirects to Morphology (linguistics)):

The popupFixRedirs feature allows you to hover over the morphosyntactical link, click Redirects, and change the text to:

(Note: Please don't try it on this page. Popups fixes every case of a redirect where it finds it on a page, and so this text will make no sense if you "fix" it.)

Myth
The new wikitext avoids the redirect and goes "straight" to the "right page", so it must be "better" on the Wikipedia servers, and you're performing a great service by "fixing" the "bad" link, right?

Fact
Unfortunately, probably not. Here's why:
 * When someone clicks on a redirected link (say, morphosyntax), the MediaWiki software performs SQL queries to locate the destination page. That query includes a check as to whether the requested destination is a redirect (every page includes a database field called page.page_is_redirect), and if so, it returns the location of the redirected page.  In other words, while it's technically true that it's more expensive to follow a redirect than a direct link, it's only the tiniest hair more expensive.  (For those who care, the actual number on a 2006 version of MediaWiki was 44 SELECT statements to follow a direct link versus 48 to follow a redirect.)
 * When you use Popups to fix a redirect, MediaWiki has to do more SQL queries than it takes to follow the redirect, but it also has to do several database transactions and writes. (To be exact, 8 transactions with 64 SELECTs, 10 UPDATEs, 4 INSERTs, and 2 DELETEs, on that same version of MediaWiki.)
 * Database updates – that is, write operations – are several orders of magnitude more expensive on the servers than database queries — that is, read operations. A benchmark ran in January 2006 showed that fixing a redirect is approximately ten thousand times more expensive for the server than following that redirect.

In other words, readers of Wikipedia would have to use a redirect link about 10,000 times before it would be worthwhile to replace that link with a direct link. In any case, Don't worry about performance discourages Wikipedians from worrying about performance, so you shouldn't be trying to fix a redirect to reduce the load on the servers in the first place.

There is a final, perhaps more important, reason not to fix many redirects: The redirect page might be about another but related topic from the one redirected to, and someone might want to create the page in the future; such a page is a redirect with possibilities. When such a page is created, "fixed" redirects will point to an incorrect (or less precise) page.