Fix for Chrome Accept headers

[Michal Fojtik]

Hi everyone,

This little patch will fix Chrome behavior, which I discovered recently.
When you navigate to API page, you will get XML version of API.

This is because Chrome sending 'application/xml' mime type before
'text/html' (which is weird but standard correct).
Unfortunately, we are lacking 'q' parameter handling in 'respond_to'
plugin to handle this correctly...

[PATCH core] Fix for Chrome Accept headers

[Michal Fojtik]

---
 server/libexec/lib/sinatra/respond_to.rb |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/server/libexec/lib/sinatra/respond_to.rb b/server/libexec/lib/sinatra/respond_to.rb
index 8e54bda..a79e099 100644
--- a/server/libexec/lib/sinatra/respond_to.rb
+++ b/server/libexec/lib/sinatra/respond_to.rb
@@ -237,6 +237,11 @@ module Sinatra
       def match_accept_type(mime_types, format)
         selected = []
         accepted_types = mime_types.map {|type| Regexp.escape(type).gsub(/\*/,'.*') }
+        # Fix for Chrome:
+        if accepted_types.include?('text/html') and accepted_types.include?('application/xml')
+          accepted_types.delete('text/html')
+          accepted_types.insert(0, 'text/html')
+        end
         accepted_types.each do |at|
           format.each do |fmt, ht, handler|
             (selected = [fmt, ht, handler]) and break if ht.match(at)

[PATCH core] Fix for Chrome Accept headers

[Chris Lalancette]

On 02/22/2010 07:34 AM, [e-mail] wrote:
> server/libexec/lib/sinatra/respond_to.rb |    5 +++++
>  1 files changed, 5 insertions(+), 0 deletions(-)
>
> diff --git a/server/libexec/lib/sinatra/respond_to.rb b/server/libexec/lib/sinatra/respond_to.rb
> index 8e54bda..a79e099 100644
> --- a/server/libexec/lib/sinatra/respond_to.rb
> +++ b/server/libexec/lib/sinatra/respond_to.rb
> @@ -237,6 +237,11 @@ module Sinatra
>        def match_accept_type(mime_types, format)
>          selected = []
>          accepted_types = mime_types.map {|type| Regexp.escape(type).gsub(/\*/,'.*') }
>    # Fix for Chrome:
>    if accepted_types.include?('text/html') and accepted_types.include?('application/xml')
>      accepted_types.delete('text/html')
>      accepted_types.insert(0, 'text/html')
>    end
>    accepted_types.each do |at|
>      format.each do |fmt, ht, handler|
>        (selected = [fmt, ht, handler]) and break if ht.match(at)

Yeah, I ran into this too.  While it is early on, and we aren't guaranteeing
backwards-compatibility yet, the cost of doing this is close to zero.

ACK

[PATCH core] Fix for Chrome Accept headers

[Michal Fojtik]

Please ignore this patch, it will break (/api.xml) functionality in Firefox
browser.

On 22/02/10 13:34 +0100, [e-mail] wrote:
> server/libexec/lib/sinatra/respond_to.rb |    5 +++++
> 1 files changed, 5 insertions(+), 0 deletions(-)
>
> diff --git a/server/libexec/lib/sinatra/respond_to.rb b/server/libexec/lib/sinatra/respond_to.rb
> index 8e54bda..a79e099 100644
> --- a/server/libexec/lib/sinatra/respond_to.rb
> +++ b/server/libexec/lib/sinatra/respond_to.rb
> @@ -237,6 +237,11 @@ module Sinatra
>       def match_accept_type(mime_types, format)
>         selected = []
>         accepted_types = mime_types.map {|type| Regexp.escape(type).gsub(/\*/,'.*') }
>    # Fix for Chrome:
>    if accepted_types.include?('text/html') and accepted_types.include?('application/xml')
>      accepted_types.delete('text/html')
>      accepted_types.insert(0, 'text/html')
>    end
>   accepted_types.each do |at|
>     format.each do |fmt, ht, handler|
>       (selected = [fmt, ht, handler]) and break if ht.match(at)
>
> --
> 1.6.6.2

[David Lutterkort]

On Mon, 2010-02-22 at 13:34 +0100, [e-mail] wrote:
> This is because Chrome sending 'application/xml' mime type before
> 'text/html' (which is weird but standard correct).
> Unfortunately, we are lacking 'q' parameter handling in 'respond_to'
> plugin to handle this correctly...

We do .. the rack/accept_media_types gem sorts according to 'q' value.
What's the Accept header that Chrome sends ?

David

[Michal Fojtik]

On 22/02/10 14:34 -0800, David Lutterkort wrote:
> On Mon, 2010-02-22 at 13:34 +0100, [e-mail] wrote:
> > This is because Chrome sending 'application/xml' mime type before
> > 'text/html' (which is weird but standard correct).
> > Unfortunately, we are lacking 'q' parameter handling in 'respond_to'
> > plugin to handle this correctly...
>
> We do .. the rack/accept_media_types gem sorts according to 'q' value.
> What's the Accept header that Chrome sends ?

Sure, rack/accept_media_types handles this properly (IMHO), but in some
weird way it doesn't work in Chrome (Chrome sends application/xml always
first, either if you request explicitly text/html (by .html).

Chrome headers:

If I request /api, Chrome sends:

"application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5"

If I request /api.xml, Chrome sends:

"application/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5"

[David Lutterkort]

On Tue, 2010-02-23 at 13:35 +0100, Michal Fojtik wrote:
> On 22/02/10 14:34 -0800, David Lutterkort wrote:
> > On Mon, 2010-02-22 at 13:34 +0100, [e-mail] wrote:
> > > This is because Chrome sending 'application/xml' mime type before
> > > 'text/html' (which is weird but standard correct).
> > > Unfortunately, we are lacking 'q' parameter handling in 'respond_to'
> > > plugin to handle this correctly...
> >
> > We do .. the rack/accept_media_types gem sorts according to 'q' value.
> > What's the Accept header that Chrome sends ?
>
> Sure, rack/accept_media_types handles this properly (IMHO), but in some
> weird way it doesn't work in Chrome (Chrome sends application/xml always
> first, either if you request explicitly text/html (by .html).

You do get HTML though when you request it explicitly, right ?
> Chrome headers:
>
> If I request /api, Chrome sends:
>
> "application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5"
>
> If I request /api.xml, Chrome sends:
>
> "application/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5"

Chrome is broken. It prefers application/xml over text/html ? See [1]
for a nice rant.

David

[1] http://www.newmediacampaigns.com/page/browser-rest-http-accept-headers