5

On one of my sites I get 404s because some scripts are fetching all URLs from the start page as raw URLs, so they include the # in the URL. Normal browsers don’t ever send that part to the server, but these scripts do.

This is how a typical requests URL looks like:

/2014/how-to-manage-wordpress-multisite-imports-with-wp-cli/#comments

The # is not URL encoded.

I tried both following methods:

RedirectMatch 301 \#comments       /

and

RewriteRule #(.+)$ /? [L,R=301]

Both without success, the rules don’t catch these requests, because the # starts a comment. The referer and the user-agent fields are empty.

What should I do?

| improve this question | |
  • Since this is a "script" (not a browser) which is sending these malformed URLs, is a 404 not the correct response? – MrWhite Jan 13 '15 at 21:40
  • I was thinking that "400 Bad Request" would be the appropriate response. – Stephen Ostermiller♦ Jan 14 '15 at 2:48
  • "Normal browsers don’t ever send that part to the server, but these scripts do." - Are you absolutely sure? Just because you tell JS to request that URL doesn't mean the browser will do it. Have you checked the Network tab in Firefox/Chrome Dev Tools? – DisgruntledGoat Jan 15 '15 at 12:22
  • @DisgruntledGoat My question is about requests that I logged on the server, not something my browser sends. I don’t even know how to reproduce these requests, because wget, curl and PHP strip everything from the # on. – fuxia Jan 15 '15 at 12:30
4

From the mod_rewrite documentation you need to use the NE (no escape) flag when your rewrite rule has a hash:

RewriteRule #(.+)$ /? [L,R=301,NE]

You commented that the NE flag may only apply to the target URL and not the rewrite pattern. If that is the case, another approach would be to escape the # sign. mod_rewrite supports \x style escape sequences. The escape sequence for # would be \x23. So your rewrite rule could be:

RewriteRule \x23.+$ / [L,R=301]

If you want to test a solution, you can do so with telnet on the command line. Use the command line telnet example.com 80 to open a socket to your webserver. Then make a simple request like this:

GET /#test HTTP/1.0
Host: example.com

Followed by an extra new line.

| improve this answer | |
  • 1
    I have read that as if [NE] is for the target URL only, not for the rewrite pattern. Will try it. – fuxia Jan 13 '15 at 21:26
  • ...and, as the link suggests, flags apply to mod_rewrite, not mod_alias. However, I suspect the "fragment identifier" part is stripped from the URL that is matched by the RewriteRule pattern (just like the query string). You could try RewriteCond and test against %{QUERY_STRING} or %{THE_REQUEST}? – MrWhite Jan 13 '15 at 21:39
  • You are right, w3d. I put the flag on the wrong rule. I've edited my answer now. – Stephen Ostermiller♦ Jan 13 '15 at 23:44
  • \x23 doesn’t seem to work too. I will contact the hoster and report back when I get some results. – fuxia Jan 20 '15 at 20:38
  • @toscho How are you testing this? Or are you just waiting for one of these "invalid" requests coming through? You say the "# is not URL encoded" (presumably looking at your logs), however, an unencoded # in the URL is strictly invalid, so even if you are able to match the # it would not be correct to return a valid response (but also, who/what are you returning a response to?). You say your server is responding with a 404... what URL is your script seeing in your (custom) 404; if your custom 404 is being called at all? – MrWhite Jan 21 '15 at 18:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.