If something goes wrong with my site, I like to know of it. My top three problems are:
This article covers the second topic.
One thing I’m curious about is hits to non-existent pages (404s) on my site. I usually get 404s because:
- I renamed the page
- Someone typed a wrong URL
- Someone followed a wrong link
Find the 404
The first problem is to know when someone gets a 404. I’ve seen sites that tell you to contact the administrator in case of a 404. That’s crazy. The administrator should automatically detect of 404s! Almost every web server provides this facility.
The real issue is attention. I receive 700 404s a day. That’s too much to manually inspect. And most of these are not for proper web pages, but for images (for example, almost all my 404s used to be for browsers requesting favicon.ico) or weird MS Office files.
I’m interested in a small subset of 404 errors. Those that hit a web page, not support files. And those requested by a human, not a search engine or a program.
<script>(new Image()).src = "/log.pl";</script>
Every time this code runs, it loads a new image. The source of the image is a Perl script, log.pl. Every time log.pl is accessed, it logs the URL from which it was called. I’m reasonably guaranteed that these are web pages a human tried to access.
Point to the right page
Sometimes, the change happens because I changed the URLs. I keep fiddling with the site structure. Someone would have links to an old page that I’ve renamed. I may not even know that. Even if I did, they can’t be bothered with requests to change the link. So I’ve got to handle it.
The quickest way, I find, is to use Apache’s mod_rewrite. You can simply redirect the old URL to the new URL. For example, I used to have a link to
/calvin.html which I now point to
/calvinandhobbes.html. That becomes a simple line on my .htaccess file:
RewriteRule calvin.html calvinandhobbes.html
I don’t do this for every site restructuring, though. I just restructure, wait for someone to request a wrong page, and when my 404 error log warns me, I create a line in the .htaccess. It keeps the redirections down to a minimum, and only for those links that are actually visited.
Be flexible with the URL structure
Sometimes people type in a wrong link. Often, these are unintentional. Here are some common misspellings for my Hindi songs search.
s-anand.net/hindi/ s-anand.net/Hindi s-anand.net/hiundi
Occasionally, people are exploring the structure of my site:
s-anand.net/excel s-anand.net/music s-anand.net/hits
I need to decide what to do with both cases. For the former, sometimes my URL structure is too restrictive. I mean, why should someone have to remember to type
/hindi instead of
/hindi/? Who cares about case? Who cares about a trailing slash?
In such cases, I map all the variants to the right URL using mod_rewrite. For example, typing
s-anand.net/HiNDi (with or without caps, with or without a slash at the end) will still take you to the right page.
As I keep discovering new mis-spellings, I take a call on whether to add it or not. The decision is usually based on volume. If two people make the same spelling mistake in a day, I almost certainly add the variant. Most of the time, it’s just typing errors like
/hiundi which isn’t repeated oftener than once a month.
To handle the exploratory URLs, and people following wrong links, I’ve turned my custom 404 page into a search engine.
For example, when someone types
s-anand.net/excel, I know they’re searching for Excel. So I just do a Google Custom Search within my site for “excel” — that is, anything following the URL.
It’s a bit more complex than that, actually. I do a bit of tweaking to the URL, like convert punctuations (underscore, hyphen, full-stop, etc.) to spaces, remove common suffixes (.html, .htm) and ignore numbers. Quite often, it matches something on my site that they’re looking for. If not, ideally, I ought to try for various alternatives and subsets of the original search string to figure out a good match. But given that the number of mismatches is down to about one a day, I’m fairly comfortable right now.
What this means, incidentally, is that my site is, by default, a search engine for itself. To search for movie-related stuff on my site, just type
s-anand.net/movie and you get a search of the word “movie” on my site. (Sort of like on a9.com, where searching for
a9.com/keyword does a search on the keyword.)