I've relied on tools like Multiple IE and IETester to test sites in IE6 and 7. While these tools are immensely useful, I sometimes wonder how accurate they are. Do you know of any resources that list known differences between these browser emulators (for lack of a better word) and the 'real deal' standalone browsers? Have you experienced any issues/oddities with IETester, Multiple IE, or similar tools?
Both IETester and multiple IEs have problems and inconsistencies when using IE6, neither of them give a 100% accurate IE6 experience. They are also quite lacking when it comes to developer tools.
I have been using IECollection for months now and it is much better than either IETester or multiple IEs and gives a much truer representation of legacy IE versions. It also ships with the proper IEDeveloper Toolbar, which while not as good as Firebug is still a massive improvement on Debugbar and other third party DOM explorers.
However, the only 100% accurate way to browser test more than one version of IE on the same machine is to run a virtual machine for each version of IE that you want to test; I did this for a while and made a stripped-down version of windows XP using nLite (approx half the size of a normal install) and used Virtual PC to have a version of XP installed just for IE6 and one installed just for IE7. Plus the proper full install for IE8.
It worked really well although I switched over to IECollection because my computer just wasn't up to spec...
When we first started building browser testing at Litmus, we used MultipleIE. We found some discrepancies pretty quickly and had to move to native installs.
One example we had was IE6 rending a particular site correctly, even though in a native IE6 install it had problems. It looked as though IE6 was using some IE7 layout logic (IE7 was the native browser on the machine).
There was also an issue with conditional comments, but that was fairly straight-forward to fix.
MultipleIE (and IETester, AFAIK), use a hack called DLL redirection, which was an early work around for DLL naming conflicts. They rely on the browser detecting a local copy of the DLL and using that instead of the registered one for that machine. I can only assume the discrepancies we saw were due to some DLLs not being redirected correctly. But that's just a guess.
I have found the exact same problems as Matthew mentioned and I have ended up installing virtual machines with nearly default installs of each browser as necessary. The only changes I make are bringing the browsers up to their most current version of the major version (such as IE6 is at 6.0.2900.5512.xpsp_sp3_gdr.090206-1234, but not upgraded to IE7 or 8).
This also works for different versions other browsers, such as Firefox, Safari (on Windows at least), Chrome or Opera.