Apple pushes Safari browser for Windows

[Darth Wong]

Has anyone else noticed that in the last day or so, Quicktime for Windows has started asking people to install Safari in its software update window? That's an interesting way of quickly distributing Safari on a shitload of Windows PCs: just tell all the users that it's an upgrade for Quicktime, and most of them will just hit the "OK" button.

PS. Well, it looks nice enough. Text looks a bit fuzzy, like it's anti-aliased. It reminds me of Konqueror on Linux, which shouldn't be surprising since I hear Safari copied KHTML from the Konqueror project. But since Konqueror is only available on Linux, I suppose Safari on Windows is the closest I'm going to come to Konqueror for Windows. I'll have to try using it for a while to see how it performs for me.

[Darth Wong]

Hmmm, posting from Safari now. Memory use is twice as large as Firefox (~60 MB as opposed to ~25MB), although that's pretty inconsequential with modern computers that usually have 1GB or more of RAM. Mine has 1.5GB. Subjectively, it actually feels faster than Firefox for some reason, although that could be my imagination. The little animated widgets in the interface are cute.

[phongn]

Safari's antialiasing is designed to preserve the look of the font even if it makes them appear somewhat blurry; Linux and Windows prefer to increase legibility in low-resolution displays (like the vast majority of all monitors). Joel on Software has a good summary of the issue.

The engine behind Safari, WebKit, was originally forked off from KHTML; work is being done to reunify the two engines. IIRC, WebKit is faster than Gecko (Firefox's engine) right now, but speed and memory improvements are promised for Firefox 3.

As for why it's being bundled - Apple probably is probably trying to increase marketshare of Safari (it's also the browser for the iPhone) and use it for the iTunes Store (which uses some custom renderer right now, IIRC.)

[Durandal]

Has anyone else noticed that in the last day or so, Quicktime for Windows has started asking people to install Safari in its software update window? That's an interesting way of quickly distributing Safari on a shitload of Windows PCs: just tell all the users that it's an upgrade for Quicktime, and most of them will just hit the "OK" button.

Yes, and there's a somewhat justifiable shitstorm about it.

PS. Well, it looks nice enough. Text looks a bit fuzzy, like it's anti-aliased.

It is, but it's using Mac OS X's font rendering as opposed to the native Windows rendering. Official support for ClearType will be coming soon. In the mean time, if you want Safari's fonts to look like the rest of the OS, see here.

Unless you're using Windows 2000.

It reminds me of Konqueror on Linux, which shouldn't be surprising since I hear Safari copied KHTML from the Konqueror project. But since Konqueror is only available on Linux, I suppose Safari on Windows is the closest I'm going to come to Konqueror for Windows. I'll have to try using it for a while to see how it performs for me.

WebKit started as a KHTML fork, yes. For a while, the two traded fixes back and forth. A while ago, the KHTML people decided to merge with WebKit, since it exposes nicer interfaces, is already pretty widely-used outside Mac OS X and has a good embedded presence. So future versions of Konquerer will very likely be using WebKit.

As for why it's being bundled - Apple probably is probably trying to increase marketshare of Safari (it's also the browser for the iPhone) and use it for the iTunes Store (which uses some custom renderer right now, IIRC.)

Yes, iTunes does use its own little thing to display the music store, but I doubt that had anything to do with the decision to push new software through a program called "Software Update". If the iTunes people wanted to transition to using WebKit for the store, they could just throw WebKit in the iTunes update.

[Praxis]

Hmmm, posting from Safari now. Memory use is twice as large as Firefox (~60 MB as opposed to ~25MB), although that's pretty inconsequential with modern computers that usually have 1GB or more of RAM. Mine has 1.5GB. Subjectively, it actually feels faster than Firefox for some reason, although that could be my imagination. The little animated widgets in the interface are cute.

Apple claims that Safari's rendering engine is faster than anything else out there, and in my experience on both Macs and PCs it's usually true. It feels snappier because it loads HTML faster. Though probably not by the margin Apple claims.

I haven't tried the Windows version of Safari lately. The old beta was horribly buggy so I uninstalled...maybe I should give it another shot, but I'm kind of comfortable in FireFox. I do use Safari on my Mac more because of the snapiness though.

Apple pushing it through Quicktime updates is clever, and rather underhanded. Seems more like a Microsoft move. I dunno what to make of it. On the one hand, it's been proven that many idiots don't even realize that alternative browsers are available and won't use anything except what is stuck in front of them, so I'd like to think this'll encourage people to try alternative browsers- but I hate the way it is being done. It should not be checked by default in the Software Update for iTunes or QuickTime, just there as an option.

I expect as desktop support I'm going to find a lot of people at work installing it on accident alongside QuickTime and I'll have to explain why some of our poorly coded intranet pages that only work on IE aren't working for them.

[Durandal]

Safari's antialiasing is designed to preserve the look of the font even if it makes them appear somewhat blurry; Linux and Windows prefer to increase legibility in low-resolution displays (like the vast majority of all monitors). Joel on Software has a good summary of the issue.

It's not that cut and dry though. Sharper text definitely does aid readability up to a certain point, but font designers typically design fonts a certain way to increase legibility. While individual letters may be a little fuzzier, your brain sees the whole word, as opposed to the individual letters, and recognizes the word as a shape that has a certain outline. Things like ascenders, descenders and serifs are part of this outline. (It takes more effort for you to pick out a certain letter in a word than to simply read the entire word.)

All of these things are modified by ClearType when they are forced to align to the pixel grid. This can counteract legibility, since it second-guesses the font designer. Quartz will preserve them as much as it can.

ClearType also only forces glyphs to vertical pixel boundaries. On horizontal boundaries, pixels can fall on sub-pixels. If they didn't, a font's kerning would be mangled beyond belief. This is why ClearType-rendered glyphs have a vertically stretched appearance (and frankly, makes some fonts look just downright ugly). This will also mess with ligatures, which, like serifs, are meant to bind a word together and aid readability.

As to the magnitude of the impact, that's up for debate. But simply saying "Sharper glyphs are always make more readable words" isn't exactly true. If sharpness was the only determining factor in readability, we wouldn't be anti-aliasing fonts at all.

[Dominus Atheos]

Hmmm, posting from Safari now. Memory use is twice as large as Firefox (~60 MB as opposed to ~25MB), although that's pretty inconsequential with modern computers that usually have 1GB or more of RAM. Mine has 1.5GB. Subjectively, it actually feels faster than Firefox for some reason, although that could be my imagination. The little animated widgets in the interface are cute.

Apple claims that Safari's rendering engine is faster than anything else out there, and in my experience on both Macs and PCs it's usually true. It feels snappier because it loads HTML faster. Though probably not by the margin Apple claims.

Oh, it's goddamn fast.



As you can see, it's neck and neck with the latest beta for firefox 3. It's 4 times faster then firefox 2, and 7 times as fast as internet explorer 7.

[bilateralrope]

Hmmm, posting from Safari now. Memory use is twice as large as Firefox (~60 MB as opposed to ~25MB), although that's pretty inconsequential with modern computers that usually have 1GB or more of RAM. Mine has 1.5GB. Subjectively, it actually feels faster than Firefox for some reason, although that could be my imagination. The little animated widgets in the interface are cute.

Apple claims that Safari's rendering engine is faster than anything else out there, and in my experience on both Macs and PCs it's usually true. It feels snappier because it loads HTML faster. Though probably not by the margin Apple claims.

Does a faster rendering engine mean much when the bottleneck is how fast you can download the page ?

[Durandal]

Unless you're on an extremely slow connection or are fetching utterly gigantic pages, you probably will be able to tell the difference. Remember that HTML and CSS are just text. You can usually get them pretty quickly. But parsing, calculating layout and displaying are all things that can easily take longer than actually fetching the data. This is why modern rendering engines defer layout nodes containing image data until the image is actually fetched, because re-rendering is expensive.

And JavaScript is a different thing entirely. You'll almost certainly notice the difference between a slow JavaScript engine and a fast one.

[phongn]

All of these things are modified by ClearType when they are forced to align to the pixel grid. This can counteract legibility, since it second-guesses the font designer. Quartz will preserve them as much as it can.

Right, but the font designer (usually) is working on the assumption that his work is for print. Those constraints may not work properly when someone is viewing at a 96 ppi screen instead of 300 dpi print. Pure screen fonts get around this problem since those fonts assume low-resolution display, but most fonts don't.

ClearType also only forces glyphs to vertical pixel boundaries. On horizontal boundaries, pixels can fall on sub-pixels. If they didn't, a font's kerning would be mangled beyond belief. This is why ClearType-rendered glyphs have a vertically stretched appearance (and frankly, makes some fonts look just downright ugly). This will also mess with ligatures, which, like serifs, are meant to bind a word together and aid readability.

Well, even if they do look ugly, what if it improves legibility? Until we get to high-resolution displays, there remains a tradeoff here. As for ligatures and other advanced font features, most applications and documents don't even use them.

As to the magnitude of the impact, that's up for debate. But simply saying "Sharper glyphs are always make more readable words" isn't exactly true. If sharpness was the only determining factor in readability, we wouldn't be anti-aliasing fonts at all.

No, but it is a major factor, isn't it? And AA is actually disabled at smaller font size due to legibility as well (at least on Windows). There's also the behavior of that AA on dark backgrounds with light text: legibility seems to be reduced, there.

[Durandal]

Right, but the font designer (usually) is working on the assumption that his work is for print. Those constraints may not work properly when someone is viewing at a 96 ppi screen instead of 300 dpi print. Pure screen fonts get around this problem since those fonts assume low-resolution display, but most fonts don't.

That means that there's a reason to render accurately, so that word forms and shapes are preserved. Take, for example, a font where p's and q's are only differentiated by the placement of the descender. ClearType might clip the descender to make it fit the pixel boundary, which, if the text is underlined would make it difficult to distinguish between the two.

Well, even if they do look ugly, what if it improves legibility? Until we get to high-resolution displays, there remains a tradeoff here. As for ligatures and other advanced font features, most applications and documents don't even use them.

Ligatures are merely an example. And yes, it is a trade-off. But on low-resolution displays, there is not just one side that has a claim solely on legibility, which is my point. Accuracy can, in many cases, improve legibility because of the way the brain perceives words. Word form perception is much more nuanced than Joel makes it out to be on his blog.

To my eyes, ClearType is actually rather harsh in its treatment of fonts. It tends to give a lot of weight to serifs while being content to give a horizontal line a single pixel row. Because of this, a word can lose its symmetry in its baseline and midline (i.e. the horizontal bar in 'A' does not match match up with the one on a 't').

No, but it is a major factor, isn't it? And AA is actually disabled at smaller font size due to legibility as well (at least on Windows).

On Mac OS X, you can set the threshold for anti-aliasing.

And at small font sizes, accuracy can be important too, because rendering the font without anti-aliasing discards more information about the glyph. You could easily run into a scenario where, at a small point size, an 'i' and an 'j' look very similar. If forced to align to a pixel boundary, the descender on the 'j' could be clipped to the point where you could no longer tell it was there. In such a case, accuracy would be preferred to sharpness.

I'm not saying one is any better than the other. I personally prefer Quartz, but it's a totally subjective opinion. To a lot of people, more accurate fonts are more legible than worse-looking but sharper ones. That's the point I'm trying to get across.

There's also the behavior of that AA on dark backgrounds with light text: legibility seems to be reduced, there.

It depends on which colors you choose, I guess. But, for example, the default theme on phpBB looks just fine in Safari, to me. I don't have problems reading the white text on the dark blue background.

[Darth Wong]

It depends on which colors you choose, I guess. But, for example, the default theme on phpBB looks just fine in Safari, to me. I don't have problems reading the white text on the dark blue background.

I agree; the white text is fine. It's the light blue text in BlackSoul which looks particularly fuzzy on Safari in Windows. More so than it does on Konqueror running in Linux with text anti-aliasing on.

[Xisiqomelir]

Should I be amazed that we headed down the pixel grid tangent path? Probably not. Respecting the typeface is more important by far, btw. Nothing is worse than trying to print something from Winblows and having the resultant page bear not even a passing similarity to what you saw on-screen.

[andrewgpaul]

Does this stealth download force Safari to be the default browser? If not, how will the aforementioned clueless users even notice it's there?

[phongn]

That means that there's a reason to render accurately, so that word forms and shapes are preserved. Take, for example, a font where p's and q's are only differentiated by the placement of the descender. ClearType might clip the descender to make it fit the pixel boundary, which, if the text is underlined would make it difficult to distinguish between the two.

Er, I don't think I've ever seen ClearType break any descenders that badly (then again, I tend to use fonts designed for the screen wherever possible if something isn't going out to print).

Because of this, a word can lose its symmetry in its baseline and midline (i.e. the horizontal bar in 'A' does not match match up with the one on a 't').

Under what font? I don't think I've ever experienced baseline/midline issues.

On Mac OS X, you can set the threshold for anti-aliasing.

And at small font sizes, accuracy can be important too, because rendering the font without anti-aliasing discards more information about the glyph. You could easily run into a scenario where, at a small point size, an 'i' and an 'j' look very similar. If forced to align to a pixel boundary, the descender on the 'j' could be clipped to the point where you could no longer tell it was there. In such a case, accuracy would be preferred to sharpness.

In this case the letters tend to bleed into each other, which throws me off a bit.

I'm not saying one is any better than the other. I personally prefer Quartz, but it's a totally subjective opinion. To a lot of people, more accurate fonts are more legible than worse-looking but sharper ones. That's the point I'm trying to get across.

True enough, but what's the common case? I wonder if any actual studies have been done regarding this (it would seem to be a reasonable one for human-computer interactions).

There's also the behavior of that AA on dark backgrounds with light text: legibility seems to be reduced, there.

It depends on which colors you choose, I guess. But, for example, the default theme on phpBB looks just fine in Safari, to me. I don't have problems reading the white text on the dark blue background.

I have some issues there (as well as, for example, on boards that have themes like Spacebattles), even on the "Light" setting the text looks exceedingly fuzzy to me.

Respecting the typeface is more important by far, btw. Nothing is worse than trying to print something from Winblows and having the resultant page bear not even a passing similarity to what you saw on-screen.

You're grossly exaggerating what Windows' font rendering does, and if screen-to-print accuracy is that important, you should be using a DTP application anyways (or at least do a PDF proof).

Does this stealth download force Safari to be the default browser? If not, how will the aforementioned clueless users even notice it's there?

It doesn't automatically set itself as default but it does automatically place an icon on the desktop.

[Dahak]

Well, comparing Safari, IE 7 and Opera 9, I prefer Opera. Maybe it's because it's the one I am most used to, but to me, it looks sharper.
IE 7 and Safari are much to mushy for my taste, with Safari feeling like a tiny bit worse than IE7. But I have to give Safari a speed advantage over the current Opera...

[TimothyC]

Does this stealth download force Safari to be the default browser? If not, how will the aforementioned clueless users even notice it's there?

Not exactly. The option to launch Safari is checked after install, and the always check to set Safari as default is also checked.


EDIT: I'm not sure I like the way the fonts are displayed, but I will admit that it feels much, much faster than either Fx 2.0.0.12 or Opera 9.26.

If they made it so that the GUI fit more with Windows they might actually get me away from Fx/Opera to a Safari/Opera combo.

[Drooling Iguana]

Well, this is without a question a dirty, underhanded move on Apple's part. However, getting alternative browsers on more systems will go a long way to making a standards-compliant web, so this might do more good than harm.

Is WebKit GPL? Do we run the risk of Apple gaining a dominant position in the web browser market and proceeding to break standards the way Microsoft does, or will they be forced to give any changes they make back to the community?

[phongn]

Is WebKit GPL? Do we run the risk of Apple gaining a dominant position in the web browser market and proceeding to break standards the way Microsoft does, or will they be forced to give any changes they make back to the community?

As noted on the main WebKit webpage, "WebKit is open source software with portions licensed under the LGPL and BSD licenses."

[Durandal]

Is WebKit GPL? Do we run the risk of Apple gaining a dominant position in the web browser market and proceeding to break standards the way Microsoft does, or will they be forced to give any changes they make back to the community?

WebKit is LGPL and BSD-licensed. And with David Hyatt leading the WebKit project, I don't think you have to worry about breaking standards.

And Microsoft didn't pull a move like that with IE. (They sort of did with Samba.) What they did with IE was just bury the other guy's proprietary standards with theirs.

True enough, but what's the common case? I wonder if any actual studies have been done regarding this (it would seem to be a reasonable one for human-computer interactions).

Well, according to some, there is an area of the brain dedicated to processing words as whole units. But a quick Googling gives me the impression that not all neuro-scientists agree with that. There's also a study showing that people misidentify characters less frequently in Microsoft's screen-designed fonts than in Times New Roman. But this doesn't speak to word legibility. And funnily enough, I use Consolas as my coding font, and I find it much more legible when rendered with Quartz than when rendered with ClearType, the algorithm it was designed for.

Though I do remember a study a while back showing that you can mix up the letters in the middle of word and a lot of the time, people will still be able to tell what word it is. Though this could be due to the brain auto-correcting the input before actually sending it off for processing. But it shows that word form perimeter and sharpness aren't the only factors in legibility.

But all you have to do is try reading a sentence written all in caps to understand that a word's perimeter shape is important too.

However, to my knowledge, there has never been a study directly comparing the impact of ClearType vs. Quartz on font legibility. If one was done, I suspect that it wouldn't really show much of a difference between the two, to be honest. Whatever you're more familiar with tends to be what you'll like better and rate as more readable.

[TimothyC]

Oh Durandal - I want to have the scroll-wheel-click scrolling I get almost everywhere else. That would be a bigger plus than a native GUI. Just something to pass on.

[Resinence]

*Checks*
As far as I can tell, clicking the scroll wheel in safari on OS X doesn't actually do anything either. And Apple was going for "exactly the same as native" I think. Either way, asking apple to adopt windows UI standards seems like a losing proposition...

[Durandal]

Oh Durandal - I want to have the scroll-wheel-click scrolling I get almost everywhere else. That would be a bigger plus than a native GUI. Just something to pass on.

What behavior are you referring to specifically? Is this a page-up or -down when you hold down the scroll wheel and scroll?

[TimothyC]

Oh Durandal - I want to have the scroll-wheel-click scrolling I get almost everywhere else. That would be a bigger plus than a native GUI. Just something to pass on.

What behavior are you referring to specifically? Is this a page-up or -down when you hold down the scroll wheel and scroll?

In Firefox, Opera, and IE, when I click and hold the scroll wheel, and move it north/south of the spot where I did so (while still holding it) the page scrolls at a rate determined by the distance from the point were I started. I tend to use this in content-light pages, and looking at it after a good night's sleep, it is a more important feature to me than a native GUI.

[andrewgpaul]

FYI, that also works when you click and release the scroll wheel, then move the mouse. It works horizontally too.