I have to hand it to IE these days, their performance boasts aren't just lip service. They are definitely outpacing Chrome and FF on canvas performance.
Chrome used to dither between best and second best in performance in the last 4 years, but for some time now IE10 + 11 have been the fastest browsers at rendering 2D Canvas paths/primitives/images by a huge margin.
I develop a canvas-based library and have a series of tests[1] that also serve as a (somewhat practical than usual) benchmark for how browsers compare when using canvas. Here's how long it takes to complete the tests on the big three:
Chrome had an edge when I started out with canvas stuff, now IE is almost twice as fast as Chrome.
[1] you can run the tests yourself here: http://gojs.net/latest/tests/index.html but remember that its only meaningful if you compare results on the same machine.
It's easy to do this when you only need to address a single platform. Most of that overhead is probably bound to graphics layer overhead. Microsoft tightly couples to DirectX. Chrome has to sit on top of SKIA to support Win/Linux/OSX/Android.
I bet Safari Canvas performance is also high, since they can directly bind to OSX.
Now I'm curious which video driver features they depend on for full optimizations. Running on a current MacBook Air inside VMWare Fusion, IE11 and Chrome have very similar performance:
IE11 in a VM: 24.874
Chrome in a VM: 24.137
As an aside, I'm impressed that the overhead of running in VMWare was that fast. It's within a factor of two of the best native performance:
Not really a benchmark, but anecdata, but for us at work (we're doing graph drawing, layouting, and editing using SVG for graphics in the browser) IE is consistently fastest, Chrome comes second and Firefox usually last.
SVG doesn't seem to be much of a priority for them, sadly.
I'm trying to use IE11, but the #1 issue I find are document modes.
I often find sites which set the wrong document mode (ex. IE9), and the site behaves broken in comparison with the IE11 document mode. (Basically people taking for granted that every IE is a faulty browser, and giving up completely).
Plus, having zero ability to tell IE: "Fuck the Internet, I want to use the newest document mode for everything!!!" is a really negative misfeature.
If anyone from the IE team reads this: Please, add a system policy to force a document mode unless compatibility mode is enabled.
It's a question of who gets the last word, though. Originally I thought that web devs should have the last word about IE document mode, because darn it, they own their site. After seeing so many sites that seem to be unmaintained (or maintained by uninformed devs) I am also leaning towards the user having the ultimate say.
I personally think that the configuration set by web developers should be respected by default, but the same way there is a "Compatibility Mode" that enforces a legacy document mode, there should be something that enforces the most recent document mode.
Feel free to vote up the suggestion on User Voice[1] so we see that more people want this feature and prioritize it correctly. I felt it prudent to leave out the suggested name, as much as I like it ;)
I'm curious about what kind of security stuff IE12 supports. Does it support new curves like Curve25519? HSTS? In many cases, even if Chrome and Firefox are more pro-active about modern crypto, IE ends up holding them back for years. I wish that stop being the case with the new versions.
Also, unrelated to IE, but why doesn't Windows 10 add out of the box encryption like iOS8 and Android L? (one where Microsoft doesn't get to keep the keys in its cloud - I hope that's clear).
IE12 isn't released so I doubt the feature set is fully established. The status.modern.ie site has been mentioned on HN a few times though, and it does have some info. https://status.modern.ie/?term=hsts
I'd love to see them harness the new-found underdog energy they have and develop a proper Mac/Linux port of IE that they can maintain alongside the desktop version.
I honestly do my best to test across all the modern browsers when I develop, but having to spin up a random VM running whatever version of windows they require for whatever the edge version of IE is starts to get old after a while.
If the Chrome and Firefox teams can maintain multiple architectures and release them in lock-step, it's certainly possible for Microsoft to start on the road to this process, even if it takes them a couple of years.
And would open-sourcing the browser really kill them?
Internet Explorer is a relatively simple executable that relies on other Windows DLLs to do pretty much everything that needs to be done. Check out the architecture here: http://msdn.microsoft.com/en-us/library/aa741312.aspx
Ideally just the Trident rendering engine (in MSHTML.dll) could be ported to other platforms for web developer use, but I don't know how reliant on other Windows features it is. I also don't think it's much of a priority (or at all feasible) for the IE team to re-architect their entire product, but I would love for them to prove me wrong.
> And would open-sourcing the browser really kill them?
No idea why not, but it's a different group than MS's other Open Sourcing efforts; The Dev tools people are happy to show their ASP.net mvc work[1], among other things[2], but IE falls under their Windows group, so it's a different set of people making the decisions.
Microsoft still lives and breathes with their old proprietary software licensing model. There would have to be some sort of major gain for them to port the browser outside of their ecosystem.
I can see them doing it with IE mobile though as they aren't the most popular player in the mobile market and they need to get people hooked.
If they have functioning brain they should - decouple the browser from the OS and then sue apple for anti competitiveness if the browser engine is not allowed on the ios store (MS is one of the few companies that could reasonably well pull this off)
I might be wrong, but as far as I am aware the last time Apple and Microsoft faced each other in court was when Apple sued Microsoft in 1988. When the dust settled, IE was the default Mac browser and Microsoft owned $150m of Apple stock.
I wouldn't bet on Microsoft and Apple ever suing each other again.
Not just living in the past, but in a specific slice of the past since in the era of IE3 (pre-Webkit/pre-Chrome/when Netscape was big on turning the browser into an Office Suite), IE was by far the best web browser.
That aside, while I do think modern IE is quite a good web browser and I mostly use Windows, I still use Chrome as my primary browser because it is everywhere (on my phone, on Linux, on Chromebooks) and IE is only on Windows.
It's cool (for those who don't have a choice / don't understand that they have a choice) that it's getting better, but when you release version 6 of your software that comes pre-installed on millions of computers, and that software chooses to ignore standards and do things a different way, requiring many many dev hours wasted on making compatibility layers...don't expect devs to forget quickly.
When you then go on to do the same thing with versions, 7, 8, and 9 of your software, all broken in new and exciting ways, mostly requiring different versions of your OS to test on...don't expect many people to still develop software for you browser platform. Fool me once, etc.
I'll make things compatible with IE again when they switch to WebKit or actually follow the standards.
Chrome used to dither between best and second best in performance in the last 4 years, but for some time now IE10 + 11 have been the fastest browsers at rendering 2D Canvas paths/primitives/images by a huge margin.
I develop a canvas-based library and have a series of tests[1] that also serve as a (somewhat practical than usual) benchmark for how browsers compare when using canvas. Here's how long it takes to complete the tests on the big three:
Chrome had an edge when I started out with canvas stuff, now IE is almost twice as fast as Chrome.[1] you can run the tests yourself here: http://gojs.net/latest/tests/index.html but remember that its only meaningful if you compare results on the same machine.