Content negotiation fails for some */* headers (regression)

Here is a monkey patch that you can drop into your rails initializers if you are in a hurry to fix this because you have a client with a mobile phone who really needs access. I'm fully aware this is not a proper solution but it might help someone.

module ActionDispatch
  module Http
    module MimeNegotiation

      # Patched to always accept at least HTML
      def accepts
        @env["action_dispatch.request.accepts"] ||= begin
          header = @env['HTTP_ACCEPT'].to_s.strip

          if header.empty?
            [content_mime_type]
          else
            Mime::Type.parse(header) << Mime::HTML
          end
        end
      end

    end
  end
end

Thank you for that, that does indeed seem to cover all the cases.

I've never seen a browser ask for image/ or application/ though, so we could make it easier on ourselves and assume any client that asks for a page with an HTML view can accept an HTML view. Simplifies the code and might be a bit more robust in serving clients that have an accept header that is plain/wrong. ("Do we want to be right or do we want to be pragmatic?"). Since you've gone through the trouble of creating this patch already I say being right is fine :-)

Just listing the list of changes done as part of second set of changes.

https://github.com/rails/rails/compare/c6030e8...fa2a5ae

(from [d67464a160bf2a5ef899a14cff93756aad374a6a]) Do not send id for quoting twice if the primary key is string.

[#6022 Content negotiation fails for some */* headers (regression) state:resolved] https://github.com/rails/rails/commit/d67464a160bf2a5ef899a14cff937...

I have another case for the same error that still fails with this commit

• HTTP_ACCEPT: */*;q=0.01

There's a semicolon in the string, the user agent was Mozilla/4.0 (PSP (PlayStation Portable); 2.00).

This one should fall back to html (or plain text), too. It was sent from Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1).

• HTTP_ACCEPT: text/*

can you describe what error msg you are getting and what's the setup?

On Rails 3.0.3 + Ruby 1.9.2, the same "template missing" error as Fjan described.

Now my patch has the following regex in ActionDispatch::Http::MimeNegotiation:

BROWSER_LIKE_ACCEPTS = /,\s*\*\/\*|\*\/\*\s*,|\*\/\*;|text\/\*/

however I'm not sure if it's a good idea to keep adding exceptions in this manner - probably we should just at least try to find html template rather than throwing an exception? I don't know if there's a good reason when "template missing" is useful on production.

Yes, I actually ended up using the monkey patch above that just return the HTML template if no other is available. There are just to many odd headers out there.

Keen Ejima,

The bug you mentioned with

HTTP_ACCEPT: */*;q=0.01

Is covered by #5833 Bug in Mime::Type#parse for single media range with accept params which includes a patch to fix.

J.

I think the monkey patch above isn't a bad idea actually – as a fallback. Maybe an option like this could actually be considered for inclusion? Otherwise, aren't we forever having to patch rails for every obscure accepts header found in the wild?

Does returning status code 500 even make sense for these "bad" (in ruby's eyes) requests?

I'm pretty happy with the idea "clients that know what they are doing can request alternative formats – everyone else gets html".