miscoded

navigator.appVersion

how about this:

alert(navigator.userAgent)

it does produce a different result everywhere, but it's a little too obvious

EDIT: robbertbroersma beat me to the punch

navigator.userAgent is shorter

navigator.appName

2 d4n3: navigator.appName is 'Netscape' in latest FF and Safari

try{x}catch(e){alert(e)}

Safari: ReferenceError: Can't find variable: x
Firefox: ReferenceError: x is not defined
Opera: <localhost>

[Error:
name: ReferenceError
message: Statement on line 2: Undefined variable: x
Backtrace:
Line 2 of inline#1 script in file://localhost/Users/tolmasky/Desktop/a.html
try{x}catch(e){alert(e)}

]
IE: [object Error]

For reference:

try{x}catch(e){alert(e)} --> 24 chars
alert(navigator.userAgent) ---> 26 chars
alert(navigator.appName) ---> 24 chars (but fails to be different in safari and ff)

Event better:

prompt(alert()) -->15 chars

Safari: alert that reads "undefined", hit OK -> confirm box that reads "undefined" with blank input field
Firefox: just doesn't run
Opera: *blank* alert box, hit OK -> confirm box that reads "undefined" with blank input field
IE: *blank* alert box, hit OK -> confirm box that reads "undefined" AND input field that contains the text "undefined"

Just doing alert(), prompt(), and confirm() produces one (different) pair of browsers in each case that are the same, same with alert(alert()) unfortunately (or fortunately?).

for(;;)

Konqueror/Linux: A script on this page is causing KHTML to freeze. If it continues to run, other application may become less responsive.\n\n Do you want to abort the script?
Opera/Linux: Freezes the tab in some weird way, keyboard input is ignored, uses up one CPU core, other tabs and browser UI still work
Firefox/Linux: Nothing appears to happen
IE6/WinXP (German; sorry, I don't have anything else): Ein Skript auf dieser Seite verursacht eine Verzögerung in Internet Explorer. Falls das Skript weiter ausgeführt wird, verlängern sich die Antwortzeiten des Computers möglicherweise erheblich\n\nSoll das Skript abgebrochen werden? (I'm pretty sure it doesn't mention KHTML in the English version, so it's definitely a different response ;-) )

Or, alternatively, if that counts:

;

Opera/Linux: shows Opera/Linux UI
Konqueror/Linux: shows Konqueror/Linux UI
Firefox/Linux: Shows Firefox/Linux UI
IE6/WinXP: Shows IE6/WinXP UI

or even the empty script, same result :D

Not as short, but...

(The first 2 are a stretch because 2 of the results for each appear to be the same to the user, but internally, they're different.)

Focus the input and press del
<input oninput="alert(1)">

Opera: Fires the event twice
Safari: Fires the event once
Firefox: Doesn't fire the event because the value didn't change
IE: Doesn't support the input event

Enter a char in the textarea:
<textarea oninput="alert(1)"></textarea>

Opera: Fires the event twice
Safari: input is broken on textareas
Firefox: Fires the event once
IE: Doesn't support the input event

<script>
var test = document.createElement("textarea");
test.value = "\n\r\n\r";
alert(encodeURIComponent(test.value));
</script>

Firefox: %0A%0D%0A%0D
Opera: %0D%0A%0D%0A%0D
Safari: %0A%0A%0A
IE: %0D%0A%0D%0A

<param>
<script>alert(document.getElementsByTagName("param")[0]);</script>

Opera: object HTMLParamElement

IE 8 beta 2: object HTMLUnknownElement

FF 3: undefined (DOM isn't ready yet)

Safari: undefined (params not inisde <object> are dropped by the parser and *never* make it to the DOM)

alert(frames);

// Opera: "[object WindowCollection]";
// Firefox: "[object Window]";
// Safari: "[object DOMWindow]";
// IE: "[object]";

alert(top)

(same result as alert(frames))

2temp1abc

alert(top) returns "[object Window]" in Opera and Firefox

hmm. I think the entries that rely on toString methods outputting different things for errors/dom objects are not within the spirit of the challenge. I do not think this constitutes a compatibility problem. I.e. nothing could possibly break because of that.

setTimeout(alert,0)

19 bytes

IE6: no alert
FF3: semi-random small integer (see http://bit.ly/2cQMwB )
Opera 9.6: empty string
Safari 3.1: undefined

Less string cast trivialities, more real browser quirks.

For different *numeric* results in every browser, try this:

var x=Date; alert(new x()/new x()/new x()/new x()/new x());

It has also nothing with the core/engine functionality, but it works, though.

//a in FF
//does nothing in Safari
//returns undefined in Opera
//Dunno about IE but probably different
alert(/a/[-1])

btw it sucks you have to register to comment
Nice blog though I read it all the time

x()

Exception handling is a serious x-browser nightmare. This scratches the surface. Did I win?

FF:
message = x is not defined
fileName = http://www.plexode.com/cgi-bin/eval.py
lineNumber = 91
stack = calc()@http://www.plexode.com/cgi-bin/eval.py:91
calc()@http://www.plexode.com/cgi-bin/eval.py:91
once()@http://www.plexode.com/cgi-bin/eval.py:72
onclick([object MouseEvent])@http://www.plexode.com/cgi-bin/eval.py:1
name = ReferenceError

IE:
name = TypeError
message = Object Expected
number = -2146823281
description = Object expected

Safari:
message = Can't find variable: x
line = 1
sourceId = 39
name = ReferenceError

Opera:
message = Statement on line 1: Undefined variable: x
Backtrace:
Line 1 of eval script

x()
Line 37 of inline#1 script in http://www.plexode.com/cgi-bin/eval.py: In function calc

out.innerHTML = textToHtml(eval(expr).toString());
Line 18 of inline#1 script in http://www.plexode.com/cgi-bin/eval.py: In function once

calc();
Line 1 of function script

once()
...
opera#sourceloc = 1
stacktrace = n/a; see 'opera:config#UserPrefs|Exceptions Have Stacktrace'

How do plugins threaten open standards? Most plugins for Firefox add to the browsers ui and don't actually touch the content. When they do its usually via a javascript, yet it has the ability to store local info and provide a nice ui to control it.

Anonymous writes:

Wow, great post! Very interesting comments!
I always had problems with the regular expressions across browsers. Now the holes are mostly patched up but generally each would do basic operations slightly differently when you tried to use regular expressions in them.

Another thing I've had fun with is stylesheets. Try including some cross-domain javascript and then have that call document.write() to add a stylesheet. I know that you shouldn't do it but I have once or twice.

After that put in some style elements sing actual markup and specify completely different and contradictory rules. I can't remeber the details but I remember a situation where Webkit would ignore the javascrpt CSS, Gecko would let you use it but also let you override it, Trident would not let you use CSS to override the javascript-included stylesheet, even if you used !important (or maybe because of it, you know IE. I'm talking back in IE6 here) and Opera would error out and pretend that there was no DOM available to javascript, and throw up some weird debugging alert error message.

I can't provide example code because I've since learned what I was doing it wrong and how to do it properly (sorry).
...
I have had some great fun working across frames, too...



(And if you're cheating with markup and using specific browser versions, try using non w3c dom in applciation/xhtml+xml. Each browser shows a different level of support for the actual javascript and IE simply spits the whole page right out because it is too pathetic to even display a well-formed web page from 1999)