Mac Web browsers

I was going to complain about this PC World article for missing most of the interesting Mac Web browsers, but I see she's going to handle them in a later column.

Personally, I keep switching back and forth between Safari and Camino; each one has features that I miss in the other. (Firefox fans can rest assured that I've tried Firefox and often use it on Linux; I like it, but the Mac version still feels like the creditable but slightly ragged port of non-Mac software that it is. Camino is in some ways the real Mac counterpart of Firefox; it's a lean Mozilla-based browser that uses the native Mac OS X interface widgets.)

Safari's page rendering and interaction are generally faster at this point; Camino's Gecko-based CSS layout is probably marginally more standard, though they both do very well with modern real-world pages. Safari's tabs have close widgets, but, on the other hand, they're drawn in a style that always confuses me as to the identity of the active tab when there are only two, whereas Camino's are completely unambiguous. Safari integrates better with the system spell-check service, and is better at letting you manually override a site's dinking with the browser controls on a case-by-case basis; but Camino has a slightly more customizable toolbar, better right-click navigation and better display of image URLs (like many other modern browsers, it scales down large images and you can display them at full size by clicking; Safari still doesn't do this).

In my experience Camino does a little better with Java applets in the real world, but Safari's less likely to choke on Flash. They are both excellent at popup blocking. Safari is marginally better at displaying Unicode-based pages, but you can only see the difference on really hard cases, such as relying on auto-encoding for pages in Indian alphabets that lack adequate encoding metadata. I prefer the way Camino looks, but your preference may vary.

I guess Freed felt compelled to review Internet Explorer/Mac since it's so historically important and was once the default browser, but given that it's dead software (and, feature-wise, has been dead much longer than she implies), I wouldn't recommend that anyone who relies primarily on Mac OS X use it except to access the few remaining sites that simply won't work with anything else. (If you're running Mac OS 9, on the other hand, it's probably still your best bet.)

Another thing.

I saw a joke page somewhere that advertised a made-up browser as having tabbed browsing: the associated image just showed the window tabs in the Windows XP toolbar.

Which got me to thinking: Are the people who say "what's the big deal with tabbed browsing? I don't get it" mostly using Windows? Because Windows tends to strongly encourage the user to put apps into full-screen mode, in which case it really wouldn't matter much whether you're opening tabs or new windows.

...I've also heard it argued that there's no logical reason for tabbed browsing on Mac OS 9 (or on Mac OS X if you use one of the hacks that give it OS 9-like window stacking), because the application-layer rule for window stacking means that the browser windows themselves can effectively act as giant tabs. Of course, these "my user interface preferences are logically superior to yours" arguments rarely matter much in the real world.

Electoral votes

Of the various Web sites attempting to track state-by-state polls and compute electoral-vote projections from them, Electoral Expectations is the cleverest: instead of trying to calculate who's ahead in the close states and awarding all the votes to that candidate, the site attempts to parcel out projected votes according to a probabilistic expectation value of electoral votes from each state, based on the stated margin of sampling error. States that are neither safe Democratic nor safe Republican get their votes parceled out to the two candidates based on the estimated probability that one or the other would win that state in an election held today. The resulting number is comparable to those calculated by other sites, except that it jumps around much less from noise in individual results; its numbers tend to vary smoothly with time as opposed to other methods' noisier output.

My one criticism of this is that we know that the "margin of error" isn't the actual margin of error; it's just the statistical sampling error, excluding systematic errors resulting from poll wording or some other methodological quirk. In national polls of such things as presidential job performance and electoral preferences, it's well-known that the spread between different polls from poll systematics is typically over twice as large as the sampling error (in most cases, probably not because of any intentional biases; these systematic errors are just really hard to eliminate). Unfortunately, while we can measure the spread in systematic errors, there's no guarantee that, say, the average systematic error will be anything like zero, so there's no way of saying just what it is, and no really good way to adjust these expectation values accordingly.

And, of course, this raises obvious philosophical questions of just what these polls purport to be measuring. They typically make no claim to predict how the election will actually turn out. They often ask an "if the election were held today" question, but they're also notoriously bad at predicting voter turnout. Still, I think that if you're going to aggregate state polls, something like the Electoral Expectation method is the way to do it.

Voter registration

If you're an American and not currently registered to vote, you can get PDF registration forms here that will probably do the job. (North Dakota doesn't have voter registration, and Wyoming and New Hampshire are being difficult; here's more about NH voter registration and here's Wyoming.) Here are the deadlines (most of them in early October).

Oh, and

Speaking of Mac Web browsers, apparently OmniWeb 5 just shipped, and John Gruber has an interesting review. Back when OmniWeb was the only really OS X-like OS X browser, I actually bought a license, and tolerated its atrocious page layout. These days the layout is much better because they dropped their homegrown engine and switched to Apple's WebCore (the KHTML-based engine developed for Safari), but according to Gruber, it's an out-of-date version. The fancy features do look interesting.