Developers › A Journal from Behind the Trenches

Browser sniffing gone wrong (again): Cars.com

October 6, 2008 · by Haavard

If you open cars.com in Opera, all you will get is a bunch of jumbled characters. Is this another bug in Opera? No, it's browser sniffing again. And not just basic user agent string sniffing either.

Using Opera's built in "Mask as Firefox", which completely changes Opera's user agent string will in fact not fix the problem. You will need something which completely hides Opera from the server. I am using Proxomitron combined with Andrew Gregory's custom config file for faking browser IDs.

So even if a site is still broken when masking as Firefox, the site could actually be digging a lot deeper to identify the browser. If you want to be absolutely sure that it is not a bug in Opera, you may want to use a local proxy which rewrites HTTP requests on the fly to make sure that the server isn't doing anything it shouldn't.

To this day, browser discrimination remains probably Opera's main compatibility problem. If only sites could at least send what they are sending to Firefox, a lot of more sites would start working.

Readers' comments 50

Left under the original article, as archived; 1 page of comments could not be recovered.

  • João EirasMonday, October 6, 2008 1:20:11 PM

    What's the problem specifically ? What are they sniffing for ?
  • Remco LantingMonday, October 6, 2008 1:43:49 PM

    It's sending gzipped content without telling the browser it's gzipped. Badly configured server I guess

    [Edit] On closer inspection, this might be confusing Opera:

    Transfer-Encoding: gzip, chunked
    Vary: Accept-Encoding
    Content-Encoding: deflate
  • Robin ZalekMonday, October 6, 2008 2:29:44 PM

    Actually, just using Proximitron without any filters seems to make the page load correctly. This is very odd indeed.

    Edit: Scratch that, this problem seems to exist on a quantum level. As any attempt I make to debug it just changes the results for me.
  • Andrew GregoryMonday, October 6, 2008 4:21:28 PM

    Oh boy, I really need to update that page! Too many other things to do! :sad:

    Removing "deflate" from the "Accept-Encoding" header results in "Content-Encoding" of "gzip" instead of "deflate", which works around whatever server-side problem they have.

    I would have thought "Mask as Firefox" would work. Since the JS component of my cfg file can't have any affect (since the page isn't decoded correctly), it must be the User-Agent string. The one in my config is rather old. Hmm, I've just tried the old UA in Fiddler2 (I don't use Proxo any more) and that didn't change anything.

    Just weird.
  • Kyle BakerMonday, October 6, 2008 6:19:50 PM

    Has anyone tried contacting them yet?
  • João EirasTuesday, October 7, 2008 12:14:59 AM

    Originally posted by Andrew Gregory:

    Removing "deflate" from the "Accept-Encoding" header results in "Content-Encoding" of "gzip" instead of "deflate", which works around whatever server-side problem they have.


    How about moving deflate to the end of the header, and leaving gzip before ?
    I saw another website that broke due to the same thing.
  • InvestorTuesday, October 7, 2008 2:57:52 AM

    >To this day, browser discrimination remains probably Opera's main compatibility problem.<

    If using a Proxy eg. http://www.quickbrowsing.com/ takes care of browsersniffing, then Opera should make a deal with them... Have a *Open With Proxy* option on menu.


    ( www.cars.com opens with quickbrowsing, but not sure site works as intended, though)
  • HaavardTuesday, October 7, 2008 7:51:57 AM

    Using a proxy doesn't solve the problem. Spoofing as another browser does. I just happen to be using a local proxy to do so properly.
  • FavTuesday, October 7, 2008 9:02:17 AM

    Investor,

    There are like hundreds of thousands of those proxy sites. It's a simple script (free). So not sure why you say they should make a deal (unless its your site). I had few of them as well...
  • Unregistered userTuesday, October 7, 2008 10:34:53 AM

    Anonymous writes: No problems with Firefox 3.1 or Google Chrome 0.3.155.0(which passes Acid3 100/100 btw). Clearly Opera and Presto again is inferior in standards compliant website compatability, no surprise there. After constant headaches from using inferior Opera, switching to Firefox 3.1 and Chrome 0.3.155.0 is the best thing one can ever do. Then again I don't see much changing anyway, seeing how buggy and crappy 9.50 was until it caused plenty of people to downgrade back to 9.27 and how the Desktop Team blog is full of compatability issues complaints by Opera users that only affects Opera even with the latest 9.60 RC2, proves once again that open source is superior to close source apps. http://webkit.org/blog/280/full-pass-of-acid-3/ http://webkit.org/blog/214/introducing-squirrelfish-extreme/
  • HaavardTuesday, October 7, 2008 12:31:17 PM

    Ouch, looks like an anonymous user is a bit angry at the world today :lol:

    Anyway, I'm not sure if this is actually a browser sniffing issue after all, but I have seen quite a few sites where using Opera's masking feature is not sufficient. So the tip about using Proxomitron (or other local proxies) is still valid.
  • Andrew GregoryTuesday, October 7, 2008 3:24:26 PM

    Just for you xErath, I tried "Accept-Encoding: gzip, deflate" and nothing changed.

    And now for a little bit of interesting trivia. The "Transfer-Encoding" the site uses for this page is "gzip, chunked". According to a book I have here, "chunked" is the only encoding allowed by HTTP 1.1. It's an old book, so it could be out of date. Anyway, I changed the "TE" header Opera sends out from "deflate, gzip, chunked, identity, trailers" to "deflate, chunked, identity, trailers" (i.e. dropped gzip). The site didn't use any Transfer-Encoding at all, and it worked.

    And then while I was testing all this, Opera suddenly stopped sending HTTP 1.1 headers and started sending HTTP 1.0 requests, with no "TE" header at all. The site responded with no "Transfer-Encoding" of its own, and the page proceeded to work as expected. No idea what's happened there :confused:

    I wonder if the "gzip, chunked" is OK, and if it is, does Opera support it? Or if the site is using it correctly?
  • João EirasTuesday, October 7, 2008 7:39:03 PM

    Originally posted by anonymous:

    Clearly Opera and Presto again is inferior in standards compliant website compatability, no surprise there.


    :lol: :lol: Thanks for the lauch.
    Now xoo, go away.
  • Charles SchlossTuesday, October 7, 2008 8:00:59 PM

    I just bugged them about Opera.

    I also sent this in the message:

    Why Open the Web?

    Despite the connecting purpose of the Web, it is not entirely open to all of its users. When used correctly, HTML documents can be displayed across platforms and devices. However, many devices are excluded access to Web content.

    http://my.opera.com/community/openweb/info/
  • Charles SchlossWednesday, October 8, 2008 6:40:08 PM

    I sent a few emails back and forth to cars.com and ended up with this:

    Good afternoon,

    I'm sorry I can't provide five reasons, but one reason Cars.com does not support Opera is its improper rendering of JavaScript and CSS. In our quest to provide the greatest experience for our users, we must choose which browsers to support. We hope that you will continue to enjoy our site on a supported browser.

    Sincerely,

    Blake Washer
    Cars.com Consumer Support


    :faint:



  • Kyle BakerWednesday, October 8, 2008 7:01:54 PM

    Wow, that's a ridiculous response.
  • Charles SchlossWednesday, October 8, 2008 7:16:27 PM

    So I sent them one back:

    Are you sure because after about haveing the site open for about 5 hours the site shows back up and does not give me the srambled mess.
    Web Specifications Supported in Opera 9:
    http://www.opera.com/docs/specs/opera9/
    Browser sniffing gone wrong (again): Cars.com:
    http://my.opera.com/haavard/blog/2008/10/06/browser-sniffing-gone-wrong-again-cars-com?cid=6178770

    Charlie Schloss


    Why Open the Web?
    Despite the connecting purpose of the Web, it is not entirely open to all of its users. When used correctly, HTML documents can be displayed across platforms and devices. However, many devices are excluded access to Web content.

    http://my.opera.com/community/openweb/info/



    Now waiting for the response...
  • Charles SchlossThursday, October 9, 2008 3:42:30 AM

    Here is a weird photo:
    http://my.opera.com/Chas4/albums/showpic.dml?album=540739&picture=8439760

    its cars.com working after the site being open for about 4 hours then a page refresh :sherlock:
  • FataLThursday, October 9, 2008 5:52:54 PM

    Seems that cars.com started working! :up:
  • Charles SchlossThursday, October 9, 2008 5:58:04 PM

    After having the site open for about more than 4 hours. I try the site again I get: http://my.opera.com/Chas4/albums/showpic.dml?album=540739&picture=8443334

    :no:
  • HaavardThursday, October 9, 2008 7:02:56 PM

    It still isn't working here. But I'm not sure if my "analysis" is correct, so I might just have to hide this blog post until we're sure what's actually going on :smile:
  • Unregistered userFriday, October 10, 2008 5:04:44 PM

    Carlos Pero writes: Hi everyone. I'm Director of Technical Product Development here at Cars.com, which is a fancy title for the fact that I oversee the front-end development for the site. And I work with our Siteops team to straighten out any Web server configuration and header issues. I don't use Opera regularly, but will try to reproduce this now. I'm inside our firewall though, so I may not exactly see what you're seeing. That screenshot looks like a problem with the interpretation of the Gzipped content. I'd have to investigate precisely how and when we are Gzipping, but AFAIK we do no particular browser sniffing to simply serve our home page. (As for that customer support response, we don't actively test and QA against Opera due to resources, but at the same time I can say we endeavor to code to standards. In other words, I can't think of a reason why the home page shouldn't render fine in Opera.) If one of you more knowledgeable people figures out what precisely might help fix things on our end, I'll continue to monitor this thread.
  • HaavardFriday, October 10, 2008 6:01:00 PM

    Hi there, Carlos!

    Maybe I can get you in touch with one of our people. They've done some analysis, and I would guess that the fix is relatively simple.
  • Unregistered userFriday, October 10, 2008 9:16:42 PM

    Carlos Pero writes: I've confirmed the behavior with Opera 9.52 and hitting our live Web site. It doesn't manifest on our internal servers, because I believe we use some kind of network-level gzip service for the production Web site. If it helps, I see these 3 additional HTTP headers from www.cars.com that aren't output on our internal dev servers: Via: 1.1 Cars-XFW Vary: Accept-Encoding Content-Encoding: deflate
  • Alexey FeldgendlerSaturday, October 11, 2008 8:42:42 PM

    There is definitely some sort of server misconfiguration involved, but this doesn't seem like a case of browser sniffing.
  • Unregistered userSunday, October 19, 2008 8:13:53 PM

    Kirt writes: I contacted cars.com because I was getting the same junk in my browser (v9.6). After getting the reply, which is below, I just replied back that AutoTrader.com supports my browser so I'll be giving my business to that site. -------------- Greetings, Thank you for your response. Do have access to Safari or Mozilla. We do not support the Opera browser. Thank you, Christopher Almodovar Cars.com Consumer Support ---------------Original Message--------------- SENT: 10/18/2008 TO: [email protected] [[email protected]] CC: [email protected] SUBJECT: [feedback] I have experienced a technical issue/error message on Cars.com ===CASE INFORMATION=== Subject: I have experienced a technical issue/error message on Cars.com Name: Kirt ZIP Code: 99354 URL: http://www.cars.com/go/index.jsp Comments: Your site is coming up like crap in the Opera browser. I get nothing but a bunch of control code looking stuff. An example is what you see below... ´VmoÛ6þûWh„¶ìÄŽÝÙJ‘8Y[`M‚ÄE×}h‰¶˜H¢FRvܵÿ}w’b+íš½a‘wä½>w¼FcûÛ)þ&#65533;Ÿ&#65533;]N¦®Î!rI WïN z3Öò¼÷‡Ï;›žÁϯ§o ‚n»S#R«œÒ©ˆ=ïü‚‹œË~ð¼ÕjÕ^¶µYxÓkïžduérµl¹ÚÍvèBv܇>R„øqÊÅòø4_ÃD —i¬RÉáB®à¹H²¼³2$ÜHa‚ˆÃIî4Lbaš+Ú±WÊ@§é&#65533;i-ù[®–> têdêZn&#65533;IÕÎgNÞ;&#65533;ìA¡Zé|euk8ì¿huxÇ&#65533;ï6 ¥ÐT$Òg¡´&#65533;QùS“F¶h`fT -‡œ,È›¹6`E,áyq`–¯Uº§2Û†WÒÊîþ–k‡n/îUºÔ(©’&#034;Ei:É„‘ Ý C¡n##žyFÎhÕfÇ&#65533;M¾“ë•6¡ÙK¦Õ¬ä…¡D;i#(¼›]qÊ™<¸{ØŸ“+h=^,—5a%§ò‹—qÐUVIQ,…••š`›Å’i«?È+¢gëûêÑÿbW¸üö|z&#039;oÏ}v}yz9½a0¹¼˜ž_L1A——gW_žº‰s“=:ôáìÍuq &#65533;½Cé±ÏDŒaN1° *á³ÓÉ8d‘‘s¿*„DÏT,Ûäm“…ú®†¡ºHëÖèR$¥c@°¬ÐXLQ!‘ÑÚ‹t&#034;&#039;킊p—¸«_¸±¨–yŸMMâSäÏÐâæÞ¨Q’Úˆï×:7¶Yíû½ýÁAÏ…:ÈÄš®ïjˆ2 6‚L¸È÷F$Ya°}¶_ púÕÛé&#65533;3ˆfRÓ˜çi@ÀD|ÊôJgyÖÌMÌ rÄ1 ’&#65533;SÒîýÞü©94Ó<Žá™x®¢Ò&#65533;üÈèþ{•‚+•†zÕ&©ß8Ú\}±· ¢mVç?7>7Æ^= ®Wõ¢™× Btƒ %’ÚU¨–meD†Z뤹”áLw§Éc›‰´D4–1¢ï¿/€`w‘,ÚÚä#¯‰Ajž¡-xùÆ¢‚@½ÚÜ,å‹×j˜N¼jíMÅ&#65533;¼)¸/ÑšÞ‹ÁÁàð9æ)C¾œD2¸óç&#034;¶ˆ]&#65533;±Â-ÛæÇEʶIß%ÐNtRŠÚå»+bÞ&#65533;:Iµˆœßv8KÇñŒJU¨&#65533;bKîtæ÷;<–s‡ßݽ‘‘.)jGìx*1ï,¼&#65533;„ƒ:‡i„u0öÄ1|Ú:ë-´‡‰–÷í[›½Dè¹Á€^—²ªÈM&#65533;†Ý&#65533;¤Œ=L5þgP¥&#65533;V«–¡’vªï1mÛßX% °&ØÄy&#65533; ¹&#65533;T`7õ‹1žáÓzXûå³WJº¨¶dE{¡æ Ê`±ÁQË· ö̰éJã3$cóJDæ³ïK¸f°Ä„Xh}ÂRªn½=&#65533;dløl$¨ä3¯©…S+÷Y—÷øa&#65533;w;|xÄñ{0äÞíñî&#65533;wH:äÝ^&#65533;&#65533;&#65533;œ¼×åƒÞíòþûb¯%BÌçþƒû °úìt¿4ÑÈÀÕLöy¯Ãá°sÄaÐÿ øh¼yJñiñ L„yZéagHJ{þ•J«Ö[•¸ù†§X@s|Y&#65533;…æWÄweÀ !ùoXÐ;8 úGï- ‡ûºzàŠ¶AcÀÃ,dŸ6¢&#65533;yG#“ m>öÉÐÓ¸•h§–E÷*Ç‘“p‰£ á!V«Ò*ŠpTÖ_Y—Å«ÕåNIJv÷@Çy’>.Wj$“‚ÎŽ‹AõáÝ+ŠÕ»µž¸d)ûÖKXëê_^¥¦^Î&#65533;¿¨ìÜmNÿ&#65533;$…½4ÇA&#65533;D&Ø&#65533;ÊØ¶hByBÄ&#034;]䨶|öO•`†~M’ÿGv‰¯ÿ&#034;{#üO>9°|TÙÉí=>ó4²œ`béy&ÆRÄ*Ä$!«|IjF@ïò~TƺæÜ$8=xM«®¹‹[ wŸˆ£ÆŽøQ›E°ÍtÃ`&#65533;‰¼B}€L‰ºšÅÉ}Öþ´q›KF×Õ¼‰ûX¦ Áúðé ({!.&#65533;¼‡Jwv>V0jV ?#dÅ@Çpäˆc½ÊÉ †#}ªuÜÜc£íëþê|ºËa— ò
  • Shwetank DixitThursday, October 30, 2008 5:00:50 AM

    I contacted cars.com through their contact page as well.

    The problem caused by Opera sending a different (and perfectly valid) set of HTTP headers on reload (once it knows the server) than other browsers. The server screws up when receiving those headers. It may be a good idea to check the response of the server when recieving a request from Opera Vs other browsers. If they can make the server give the same response to Opera that it gives to other browsers, then the problem could be solved.
  • Andrew GregoryThursday, October 30, 2008 9:55:20 AM

    Just in case Carlos returns...

    I've done some further investigation using Fiddler.

    The pertinent header sent to Opera in this case is:

    Transfer-encoding: gzip, chunked

    I've found that if the "gzip" is changed to "deflate", then Opera will decode the reply. i.e. the cars.com server is not correctly indicating the compression method it has used.

    Firefox works because it does not claim to support compression in the transfer. Opera claims to support compression in the transfer because it does support it, but if the server decides to compress the data, Opera then needs to be told the correct compression method!

    It's buggy server code that has probably never been tested because no other browser claims to support transfer compression.
  • Unregistered userWednesday, November 26, 2008 9:39:24 PM

    Carlos Pero writes: Hi Andrew, et al- Your troubleshooting was very helpful in proving there was a problem. Opera appears to be sending two headers...Accept-Encoding and Transfer-Encoding, and we have two devices on our Production network that are each obeying one of the headers, and one is clobbering the output of the other. We are testing a resolution that should rectify the proper serving of content to Opera. I'll post again when we've confirmed success and deployed it to our live site. In the meantime, Happy Thanksgiving! Sincerely, -Carlos
  • Unregistered userSaturday, December 13, 2008 6:00:41 AM

    Anonymous writes: Still lots of opera users out there wanting to use cars.com... and eagerly awaiting any resolution. Any updates?
  • Charles SchlossSaturday, December 13, 2008 6:36:16 AM

    http://validator.w3.org/check for cars.com

    Errors found while checking this document as HTML 4.01 Transitional!Result: 50 Errors, 1 warning(s)

    Encoding: iso-8859-1

    Root Element: låì%Û


    I opened the page in Opera 10 Alpha 1 it opened fine then about 30 seconds later back to the scranbled code


    http://files.myopera.com/Chas4/files/Opera%2010%20alpha%201%20on%20cars.com.tiff -
  • Charles SchlossMonday, December 15, 2008 8:29:48 PM

    Here is what I get before it scrambles in Opera 10 Alpha 1:

    http://files.myopera.com/Chas4/albums/540739/Opera%2010%20alpha%201%20on%20carscom.png -

    it is something later on on the code that Opera does not like
  • Andrew GregoryWednesday, December 17, 2008 2:58:05 AM

    In all my testing, both in 9.62 and 10A1, the web page never comes up correctly, the page is always immediately scrambled.
  • Charles SchlossWednesday, December 17, 2008 3:22:03 AM

    Andrew Gregory what OS are you using? I am on Mac OS 10.4.11 Intel
  • Andrew GregoryWednesday, December 17, 2008 10:03:34 AM

    I'm using Win XP SP3.
  • Charles SchlossWednesday, December 17, 2008 8:14:46 PM

    on 9.63 I get instant scrambled code

    toward the bottom of the scrambled page is bold scabled code
    öÖé3¨^’󳑙ꣂÕA×Á™:A/ÀeL„²xß;¦÷zo,Ôg;åÇV©|½GÕ*ï#uGnë¼~õrMÈ yÝоþÜÉ 'Tûï$Œ Dƒ@´cauê@¼ûþxp–]ýÕj9‚»Pž÷9œÆéfL} ·ZÈG¹‹ÉÒ’Å 1^m+�ºJ–ÚI+†–úf&e½ J¨Ÿ·ü›N!ÈÌÔ‹U:Fƒõ‹zXÞÌ ÃÞl

    can you find anything in Opera Dragonfly in the Error Console? while at cars.com?
  • Charles SchlossWednesday, December 17, 2008 8:40:48 PM

  • Andrew GregoryThursday, December 18, 2008 5:38:54 AM

    The problem at cars.com is not Javascript related, so Dragonfly isn't particularly relevant. The problem is server-side code that doesn't correctly handle the extra compression options Opera supports. Every single page on cars.com (and pickuptrucks.com, for that matter) will be affected. Until the site fixes the problem, of course.
  • Charles SchlossWednesday, February 4, 2009 3:30:17 AM

  • Charles SchlossWednesday, February 4, 2009 7:22:36 PM

    Response:

    Case 01427743: Your inquiry to Cars.com [ ref:00D3WKx.50046sVbA:ref ]

    Good afternoon,

    Thank you for contacting us. Unfortunately, we are not able to support the Opera browser at this time. Please use one of our supported browsers, Apple Safari or Mozilla Firefox, to access Cars.com.

    Sincerely,

    Blake Washer
    Cars.com Consumer Support
  • Andrew GregoryThursday, February 5, 2009 1:14:56 AM

    Chas4: You've just reached the first-line tech support drones. You really need to reach someone like Carlos (who has posted above). cars.com are aware of the problem, and have been able to reproduce it. They just seem to be particularly slow fixing it.
  • Charles SchlossThursday, February 5, 2009 1:39:59 AM

    4 months hmmm

    funny that the drone mentioned the IE is not supported :rolleyes:
  • Unregistered userSaturday, February 13, 2010 5:08:47 AM

    Anonymous writes: a year later? by "coding to meet standards" they mean "the market". c'mon
  • Charles SchlossSaturday, February 13, 2010 5:20:47 AM

    Yep still broken more than a year later, this October will be 2 years...
  • tbutlerFriday, February 19, 2010 7:09:42 AM

    I don't know what was happening in the past, but I know why cars.com is failing now. They are compressing the html reply twice. So Opera uncompresses it and tries to render it, but it is still compressed. I figured this out by using Wireshark to capture the reply that opera sees, dumped the encoded payload, and found that after uncompressing once, it was still compressed. And uncompressing it a second time yielded the html.

    To demonstrate this concisely, I used wget with multiple --header arguments to reproduce the problem outside of Opera. There was one wrinkle with this. Wget, or at least the version I have (1.11.4) specifies HTTP/1.0 in the GET request. And the cars.com server only compresses the result if it says HTTP/1.1. Since wget doesn't provide an argument to change this, I used a binary editor to modify the executable, changing the string to 1.1. After that, I get this result:


    $ ./cget.sh
    --2010-02-19 01:51:02-- http://www.cars.com/go/index.jsp
    Resolving www.cars.com... 74.119.98.141
    Connecting to www.cars.com|74.119.98.141|:80... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 16516 (16K) [text/html]
    Saving to: `cars.reply'

    100%[==========================================================================================>] 16,516 --.-K/s in 0.08s

    2010-02-19 01:51:03 (194 KB/s) - `cars.reply' saved [16516/16516]

    $ file cars.reply
    cars.reply: gzip compressed data, from Unix
    $ mv cars.reply cars.reply.gz
    $ gunzip cars.reply.gz
    $ file cars.reply
    cars.reply: gzip compressed data, from Unix
    $ mv cars.reply cars.reply.gz
    $ gunzip cars.reply.gz
    $ file cars.reply
    cars.reply: ISO-8859 HTML document text, with very long lines, with CRLF, LF line terminators
    $


    And cget.sh is just a script to make editing the wget command a bit easier. The interesting thing is that if you drop the TE header, it fixes the problem. The reply is still compressed, but only once.

    $ cat cget.sh
    #!/usr/bin/bash

    ./wget \
    --header="Accept-Encoding: deflate, gzip, x-gzip, identity, *;q=0" \
    --header="TE: deflate, gzip, chunked, identity, trailers" \
    -O cars.reply http://www.cars.com/go/index.jsp

  • tbutlerFriday, February 19, 2010 7:28:39 AM

    I reported it to cars.com, and emphasized that it isn't browser specific. The demo can be reduced to the one shown
    below, but only if you have a version of wget that specifies HTTP/1.1 in the GET request. Note that gunzip is run on the data twice.

    $ ./wget --header="Accept-Encoding: gzip" --header="TE: gzip" -O cars.reply http://www.cars.com/
    --2010-02-19 02:24:33-- http://www.cars.com/
    Resolving www.cars.com... 74.119.98.141
    Connecting to www.cars.com|74.119.98.141|:80... connected.
    HTTP request sent, awaiting response... 302 Found
    Location: http://www.cars.com/go/index.jsp [following]
    --2010-02-19 02:24:33-- http://www.cars.com/go/index.jsp
    Reusing existing connection to www.cars.com:80.
    HTTP request sent, awaiting response... 200 OK
    Length: 16518 (16K) [text/html]
    Saving to: `cars.reply'

    100%[==========================================================================================>] 16,518 --.-K/s in 0.04s

    2010-02-19 02:24:33 (374 KB/s) - `cars.reply' saved [16518/16518]

    $ gunzip -c < cars.reply | gunzip -c > cars.reply.html
    $ file cars.reply.html
    cars.reply.html: ISO-8859 HTML document text, with very long lines, with CRLF, LF line terminators
  • John DaviesTuesday, April 27, 2010 7:53:50 PM

    Great job, tbutler, on coming up with that demo of the problem. I have one thing to add. Curl uses HTTP/1.1 by default so if you change your cget.sh script to this:
    $ cat cget.sh
    #!/usr/bin/bash
    
    /usr/bin/curl \
    --header "Accept-Encoding: deflate, gzip, x-gzip, identity, *;q=0" \
    --header "TE: deflate, gzip, chunked, identity, trailers" \
    http://www.cars.com/go/index.jsp > cars.reply
    

    then it works with an out-of-the-box binary.
  • Charles SchlossFriday, June 4, 2010 10:02:55 PM

    Found a work around if you want to see cars.com in Opera:

    It is the mobile site:

    http://m.cars.com/

    http://www.pickuptrucks.com/html/stories/international/page1.html does the same thing as the main cars.com site



    Cars.com contact page:

    http://m.cars.com/contactus.jsp
  • Charles SchlossSunday, June 6, 2010 11:33:46 PM

    I got a load of BS from cars.com:

    Case 01926656: Your Cars.com Inquiry [ ref:00D3WKx.5004At7kg:ref ]‏ [email protected] ([email protected])


    Greetings,


    Thank you for the response. While we do strive to work within all browsers, Opera currently isn’t a browser which our web site functions properly. However, we are currently looking to address this issue. Until this is resolved, we would like to suggest that you access our web site using a more industry standardized browser (i.e Internet Explorer, Mozilla FireFox, or Safari.) Please let us know if you have any questions.

    Thank you,

    Christopher Almodovar

    Cars.com Consumer Support

    We value your feedback – Please click below to answer four quick questions about your recent experience with your Cars.com Support Specialist. Thank you! http://cars.force.com/supportsurvey?caseid=5004000000At7kg


    I did mention this post in the email I sent (had to use the m.cars.com to send it tho)
  • JimWednesday, June 9, 2010 3:43:35 PM

    Perhaps you should mention that by compressing content twice it wastes server resources and slows page loading time, which is something Google is now measuring to help rank websites.

About this article

Published inA Journal from Behind the Trenches
DateOctober 6, 2008
AuthorHaavard
OriginalInternet Archive