I really, really dislike that style of numerals. It just looks like they're bouncing all around the baseline. Please don't use that, designers. Different != good.
The difference between them is not really simply a stylistic choice: text figures are basically lowercase numbers and titling figures are uppercase ones. Saying you don’t like them is like saying you don’t like lowercase letters, and that you prefer everything in uppercase. Using titling figures in running text is like typing in all caps—it’s screaming at you.
It’s not so much that the text figures are old and were abandoned, but rather that typewriters simply left them off to simplify things. And since computer keyboards were based on typewriters, they never made their way to computers either. And then they started being used exclusively in newspapers and advertising. If you look at books, though, you’ll see that they never really went away. Only now are we finally getting them back—and that’s the entire point of OpenType and it’s a good thing.
Now, the difference between text and titling figures isn’t precisely the same as upper- and lowercase, since titling figures are generally used in titles despite the fact that titles can be mixed case; and they’re also used in tables of numbers. (Although there is another dimension of figure variation—namely, whether they are proportional- or fixed-width—and fixed-width lining figures are typically used in tables so that the places line up.)
1. These are often termed old-style numerals. Nothing new under the sun here. If anything, all fixed height numerals are the newer invention. (conjecture, feel free to research this and refute, but I'd bet it's the case)
2. Research into how we read whole words indicates that letter height and the profile of entire words is critical to faster reading. SAME HEIGHT WORDS slow us down. The changes in word outline can help us along. 2009 2oo9. A poor example given that I'm simulating old-style numerals, but you get the idea. In blocks of text, not in tabular formatting, the changes in the numeral heights can help carry the eye along.
I doubt it, given that all-uppercase is the older style. http://en.wikipedia.org/wiki/Letter_case#History seems to agree:
"Originally alphabets were written entirely in capital letters, spaced between well-defined upper and lower bounds"
But feel free to research and refute.
Key takeaway (here): Yes, tons of research shows that people read lowercased words faster than they read uppercased words. It also shows that the difference disappears with practice -- turns out lowercase is more common 'in the wild'.
Text figures flow better in running text. When the numbers are part of a prose sentence and when you aren't comparing them with other numbers or doing math with them, they can be more readable (and less jarring).
Since we don't have old-style numbers in web typography today, it's understandable that they look weird to you, but they aren't a case of different- for- difference- sake.
The option to use Old Style Figures should always be applauded. As is obvious from the case of Georgia, it’s never good when you are stuck with only one or the other.
I think they're fantastic for addresses and phone numbers. Distinctiveness is a win there, since the numbers are abstract and would otherwise just run together.
The '8' in for instance "Suite 1850" is a little bouncy.
As far as it being easier to read in running text, I can't see it, but if the designer of a given work truly feels it provides that benefit, then so be it. I just would prefer for this style to not be used on the web, even though it is becoming available along with other advances in web typography. I just think it looks bad and it's hard to read.
You should read up on typography before decrying things about it. Robert Bringhurst’s The Elements of Typographic Style or Warren Chappell’s A Short History of the Printed Word are two great places to start.
But having said that: the kernel of easy- to- agree- with in your comment is, for a lot of the figures that tend to get typeset on web pages, lining figures are the more appropriate choice.
That said, when set in running text and when describing things that are, effectively, proper nouns (IP addresses, street addresses, phone numbers), old-style figures often read better and are more attractive.