It's a cute feature. It's not going to change the world, sell another billion phones, push other companies out of the market, or save anyone from serious attacks. It's probably a good idea to enable it anyway.
I think that's the key distinction here. In any given authentication scheme it's important not to have false positives (incorrectly identifying a bad guy as you) or false negatives (incorrectly identifying you as a bad guy). In this case false positives break security, false negatives break usability. However, false positives won't outright stop adoption whereas false negatives will.
With glove, dirty or too much sweat, I believe it does not work. So, it's not 'always'.
However like others, I turned it off because the performance was highly variable, and the failure mode consists of a many-seconds wait which can be extremely infuriating (even embarrassing, as as you stare blankly at your phone for 5 seconds at a party, trying to quickly get someone's number or something).
edit; not 911emergency, but casual situations of full or dirty hands..
I went to it. I leads to a special dialer. Instead of voicemail the button leads to a special emergency contact (or list). It only shows 4 inputs on top so I am guessing that is the limit so you can't dial anything but emergency services (that are 4 numbers or shorter). Then it goes back to my lock screen.
With a fingerprint unlock, I need to go to at least a little trouble to fake the fingerprint.
In neither case is the phone meaningfully protected against serious attack. Why must we have this argument? It's a cute feature. Use it.
What utter unmitigated rubbish. It is extremely unlikely that even a fully qualified CSI would be able to lift a full print from a mobile phone, let alone one that that can be reliably reproduced in the manner CCC described.
My 5 year old son was quite literally dusting for fingerprints at the local science museum last weekend. We have some shockingly high fidelity prints of both our thumbs showing all the ridges. And all we had to do was squeeze a piece of plastic. Fingerprints have even less identifying detail than faces. You've been hoodwinked by Apple's marketing, and I'm willing to bet this isn't the first time.
Yeah, not so much. The fakes didn't even pretend to be live tissue.
That doesn't seem to be the case to my knowledge. The evidence from the successful attack is that you need an excellent-quality print from one of the specific fingers that has been programmed into the phone. Some phones probably have that on them, but it appears likely that many do not.
edit: build an app, get your colleague, significant other etc touch it on any touchscreen phone or get on camera and create a 3d printed finger. 3d printing vs touchid...maybe
But putting that aside, your hypothetical app would -- using the demonstrated method -- 'lift' that excellent quality print, scan it at 2400 dpi, (clean up said print), print it on a transparency at 1200 dpi, mask it onto photosensitive PCB, develop/etch/clean the PCB, spray graphite and apply wood glue to the mold.
It might make for a slightly-more-plausible-than-normal gadget sequence in a Mission Impossible movie, but it's not much of a concern for the target market. 
 Despite what decades of shows like CSI might lead us to believe, this is not a simple or error-free process. And each mistake irrecoverably destroys the print.
 Most of that market doesn't even use a passcode today and many that do are still using surprisingly bad PINs (birthdays/anniversaries/1234)
Except when confronted with an Apple product. Then it's all "Nah bro, relax. No way could you lift a fingerprint from a glossy phone screen". :)
I'll say it for the third time. It's cute feature (like face unlock was before it). Use it and enjoy it. If you honestly think you're buying a serious security mechanism you're simply wrong.
There's security that geeks advocate for ourselves and our own implementations (often things we only have to set up and maintain infrequently) and then there's security that normals actually use (often things they have to authenticate with several times a day).
And I must have missed it, if anyone's been arguing this is a serious security mechanism. As far as I've seen, it's been lauded as (not much) better than a passcode, but, primarily, convenient enough to get people to use it instead of nothing, bringing up the relative security of a still-fairly-insecure bunch.
And you may want to re-read the discussion over the faked-print attacks. It isn't about (im)possibility. It's about the time, expertise and equipment involved and the likelihood of success being too expensive to be worthwhile for gaining access to most phones. 
And if we're wearing our "serious" security hats, I still don't see any reason to worry too much about print faking, as its core assumption is a skilled attacker who has unfettered physical access to our device, unbeknownst to us and beyond our control. And at that point, the game is already over.
 CCC themselves, with ideal source prints, had to significantly complicate their process to generate fakes that worked with a suitable consistency. So even if you think suitable source prints grow on trees, the point of significant skill, equipment, time and resources remains.
When you say it's not "a serious security mechanism", it sounds as if that's defined in some absolute terms. But if the effort to hack it is hundreds of times more difficult than the possible payoff from hacking it (which appears to be the case for nearly anybody but James Bond), then it acts as a serious security mechanism for that user's context. Literally nobody is going to make a mold of my finger to unlock my iPhone — they'd have to be absolutely insane to think that was worthwhile. So it's a serious security mechanism for me. Would it be a serious security mechanism to cover nuclear launch codes? Of course not.
You have to understand that the practice of cryptography has always had a military basis; the commercial/private use is ancillary.
So, what's "a serious security mechanism?" Presume you're a military commander during active war, whose battle plans are intercepted by an opposing nation. What is the likelihood, given the opposing nation believes your plan will lead to their complete destruction, that they'll be able to break the security in time to execute a counter-operation? A serious security mechanism is anything that reduces that likelihood.
You need to be able to sign in with the Apple ID to remove the association.
I've already done that service for another, using some auto-unlocking tools. Takes all but 5 seconds, including USB negotiation. And it even gets past sim-locks.
Of course, I don't actually use it because the face recognition is so bad, and nonexistent in the dark.
I'm ready to have a chip in my arm now.
I have no idea how finger print vs facial recognition compare in accuracy, but a decently implemented facial recognition system shouldn't be compromised by a still image.
face unlock now requires you to blink.
As they said when they unveiled the feature and people mentioned this: give them a little credit.
If I was the kind of person who was worried about someone accessing the contents of my phone, I'd simply turn off touch ID and use a long password (or spend less money on a phone that didn't have a feature I wouldn't use).
I've gone down the route of using both a long password and touch ID simply because touch ID works so reliably - I've never had to enter my password. That way someone either needs my long password or a physical copy of my fingerprint to access my device. I'd say that's much better than the 4 digit numerical code I relied on previously - which had been seen by friends and family.
Your iPhone has a picture of your fingerprint inside of it now. It's just a picture, and it's likely a very good picture at that.
What happens when I swipe your phone for a second or two, plug it into my machine, and download the high-resolution picture of your fingerprint?
Do you use a fingerprint lock at home? If so, I've just broken into your home.
Do you use a fingerprint lock for the datacenter you administer? I've just gained access.
Do you own a registered gun? How'd you like me to commit a murder with your fingerprint on it?
This kind of attack is the missing piece of my argument. When someone figures out how to do this, these issues are going to become very important very quickly.
Let's suppose that Apple introduces a feature that syncs your fingerprint across many devices. How convenient, right? Let's say that means keeping all of your fingerprints on Apple servers. Let's now suppose that, like a credit card database, an attacker is able to obtain a leaked copy of the fingerprint database of every iPhone user. The recent touchid hack shows that fingerprints can be spoofed for high-end scanners. What then?
Sure, this scenario is very unlikely. I'm totally in slippery-slope land here.
But when we choose to turn up the dial on convenience to sacrifice more security, we must be prudent, carefully considering the consequences of our intentional ignorance.
The big enterprise market is an awesome place to get a foothold in - they are not really price-sensitive and hate change. Not that Apple has any problems in that segment, but extra lock-in doesn't hurt.
Where this becomes semi-dangerous is in assuming that now your phone is ironclad and you can store whatever on it totally unprotected. The best route to safety is to make informed decisions based on your own risk-tolerance and not be a lemming.