RE: ACTION-816: Write some text for CT Appendix B.2

1)       The relevant section in the document is section 4.3.3.  I think
the way Jo edited and put it the document makes it clearer why this
particular 200 response can be treated as a 406.  (The "error" text is
more obvious instead of being buried in an HTTP response.)

 

2)       Jo added some additional text about the first response not
having a "Cache-Control: private" header.  This along with the fact that
there is no Vary header is, hopefully, pretty good evidence that there
are no specific representations for this particular user-agent.  (If the
origin server varies its response on cookies it should probably also
include a "Vary: *" header.)  Of course, who knows if a particular
origin server is going to do these things?  It's up to the CT proxy to
determine if it has enough evidence, I guess.

 

3)       Yeah, it ended up looking pretty bad.  The original email that
I sent to the list had horizontal rules between the different requests
and responses and didn't have all of the double spacing.  The mailing
list server ended mangling the content so that it was pretty hard to
read.  Interestingly, the version in the mailing list archives is much
easier to read.  No extraneous double spacing and the different requests
and responses are separated by whitespace (although the horizontal rules
are gone).  I guess it is safest to just send emails to the list using
plain text.  (So of course I'll immediately disregard my own advice
since I am not using plain text for this response.)

 

Sean

 

________________________________

From: public-bpwg-ct-request@w3.org
[mailto:public-bpwg-ct-request@w3.org] On Behalf Of Sullivan, Bryan
Sent: Thursday, July 24, 2008 12:45 AM
To: public-bpwg-ct
Subject: RE: ACTION-816: Write some text for CT Appendix B.2

 

Hi Sean,

This is a reasonable flow for this scenario. A few suggestions:

 

1) Explain how "The CT proxy determines that this 200 response equates
to a 406 and re-requests the web page from the origin server with
altered headers", or just reference the main text section where this is
explained, or say "By implementation specific means, the CT proxy..."

 

2) Do we have a caveat in the text that although the CT proxy can
"learn" ala "The CT proxy remembers that this website doesn't handle
mobile user agents properly..." that it could be mistaken in what it
thinks it has learned? Some sites already offer the ability to switch
between a mobile and desktop version (or provide distinct URLs for
that). What one user has selected may not imply correctly what the next
user should receive, since the decision on which representation to send
may be based upon information other than the URL, e.g. a cookie, which
the CT proxy is not designed to understand. If we assume that such a
flexible CP will always provide a "no-transform" indication in its
desktop content versions, then OK, but we should be explicit about that.

 

3) Indent and use a fixed width font for the snoop details; this makes
it easier to separate them from the text.

 

Thanks,

Bryan Sullivan | AT&T

________________________________

From: public-bpwg-ct-request@w3.org
[mailto:public-bpwg-ct-request@w3.org] On Behalf Of Sean Patterson
Sent: Wednesday, July 23, 2008 2:23 PM
To: public-bpwg-ct
Subject: ACTION-816: Write some text for CT Appendix B.2

Example for section B.2 of CT Guidelines:

The CT Proxy receives a request for a website that it hasn't interacted
with before:

GET / HTTP/1.1

Host: www.reallycoolnonmobilesite.com

User-Agent: MobilePhoneManufacturerInc 1000 (1.0)

Accept:
text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plai
n;q=0.8,image/png,*/*;q=0.5

Accept-Language: en-us,en;q=0.5

Accept-Encoding: gzip,deflate

Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7

The CT Proxy forwards this request to the website:

GET / HTTP/1.1

Host: www.reallycoolnonmobilesite.com

User-Agent: MobilePhoneManufacturerInc 1000 (1.0)

Accept:
text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plai
n;q=0.8,image/png,*/*;q=0.5

Accept-Language: en-us,en;q=0.5

Accept-Encoding: gzip,deflate

Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7

Via: 1.1 ronco-ct.com (Ronco CT-o-matic/7.3)

X-Forwarded-For: 10.1.1.200

The CT proxy receives the response from the origin server for the
website:

HTTP/1.1 200 OK

Date: Wed, 23 Jul 2008 19:36:25 GMT

Server: Apache

Content-Type: text/html

Content-Length: 156

<html>

<head>

<title>Unsupported browser</title>

</head>

<body>

Unsupported browser.  Please get a different one or use a CT proxy.

</body>

</head>

The CT proxy determines that this 200 response equates to a 406 and
re-requests the web page from the origin server with altered headers:

GET / HTTP/1.1

Host: www.reallycoolnonmobilesite.com

User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.16)
Gecko/20080702 Firefox/2.0.0.16 Ronco CT-o-matic/7.3

Accept: text/html;q=1.0, text/css; q=1.0, application/x-javascript;
q=1.0, text/plain;q=0.8, application/xhtml+xml;q=0.6,
application/x-httpd-php;q=0.1, */*;q=0, image/gif; q=1.0, image/jpeg;
q=1.0, image/png; q=1.0

Accept-Language: en-us,en;q=0.5

Accept-Encoding: identity;q=1.0, gzip;q=0.1, *;q=0

Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7

Via: 1.1 ronco-ct.com (Ronco CT-o-matic/7.3)

X-Forwarded-For: 10.1.1.200

X-Device-User-Agent: MobilePhoneManufacturerInc 1000 (1.0)

X-Device-Accept:
text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plai
n;q=0.8,image/png,*/*;q=0.5

X-Device-Accept-Encoding: gzip,deflate

The origin server returns the desktop version of the web page:

HTTP/1.1 200 OK

Date: Wed, 23 Jul 2008 19:36:27 GMT

Server: Apache

Content-Type: text/html

Content-Length: 29478

...29478 bytes of really cool HTML...

The CT proxy transforms this web page into content displays well on the
phone:

HTTP/1.1 200 OK

Date: Wed, 23 Jul 2008 19:36:28 GMT

Server: Ronco CT-o-matic/7.3

Content-Type: text/html

Content-Length: 14573

Warning: 214 Transformation Applied

...14573 bytes of really cool transformed HTML...

A few hours later, the CT proxy receives a request from a different user
agent for the same web page:

GET / HTTP/1.1

Host: www.reallycoolnonmobilesite.com

User-Agent: PhonesRUsInc XL 500 (3.0)

Accept:
text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plai
n;q=0.8,image/png,*/*;q=0.5

Accept-Language: en-us

Accept-Encoding: gzip,deflate

Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7

The CT proxy remembers that this website doesn't handle mobile user
agents properly and since the origin server has not indicated that it
varies on the User-Agent header (no Vary header), the CT proxy sends a
request with altered headers immediately:

GET / HTTP/1.1

Host: www.reallycoolnonmobilesite.com

User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.16)
Gecko/20080702 Firefox/2.0.0.16 Ronco CT-o-matic/7.3

Accept: text/html;q=1.0, text/css; q=1.0, application/x-javascript;
q=1.0, text/plain;q=0.8, application/xhtml+xml;q=0.6,
application/x-httpd-php;q=0.1, */*;q=0, image/gif; q=1.0, image/jpeg;
q=1.0, image/png; q=1.0

Accept-Language: en-us,en;q=0.5

Accept-Encoding: identity;q=1.0, gzip;q=0.1, *;q=0

Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7

Via: 1.1 ronco-ct.com (Ronco CT-o-matic/7.3)

X-Forwarded-For: 10.1.1.200

X-Device-User-Agent: PhonesRUsInc XL 500 (3.0)

X-Device-Accept:
text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plai
n;q=0.8,image/png,*/*;q=0.5

X-Device-Accept-Encoding: gzip,deflate

X-Device-Accept-Language: en-us

The origin server again returns the desktop version of the web page:

HTTP/1.1 200 OK

Date: Wed, 23 Jul 2008 22:45:06 GMT

Server: Apache

Content-Type: text/html

Content-Length: 29478

...29478 bytes of really cool HTML...

The CT proxy again transforms this web page into content displays well
on the phone (the transformation will be different for this phone since
the user agent is different):

HTTP/1.1 200 OK

Date: Wed, 23 Jul 2008 22:45:07 GMT

Server: Ronco CT-o-matic/7.3

Content-Type: text/html

Content-Length: 16412

Warning: 214 Transformation Applied

...16412 bytes of really cool transformed HTML...

Sean

Received on Friday, 25 July 2008 19:53:47 UTC