I'd say (in order of width) 1, 5, 3, {8, 0}, {6, 9}, {2, 7, 4}. In particular, I like the horizontals of 2, 7 and 4 to be relatively long, and I think the commonest fault is to make 3 and 5 too wide. Especially 3, which often seems to end up much too large. I like the way that in the older typefaces the lower curve on 3 and 5 is distinctly not the same shape as the circular curves on 8, 0, 6 and 9, and the horizontal of the 5 distinctly not the same length as the horizontals on 2, 4 and 7.

Personally also I like the completely unmodulated 0: the pure circle. I know people find this "stands out", and want to give it some modulation. But it seems to me to be perfectly functional and to have a certain satisfactory symbolic quality.

Numerals are good when they are tabular (i.e., all the same width), in my opinion. Otherwise, lists and tables are a pain to make.

For me the general reasoning is to keep the figures from looking out of place either with each-other or with text.
I tend to agree with Paul on the proportional categories but wonder if this is just a matter of visual adjustment to make them all feel like a set. Figures like 8 have a lot going on so they can look like a dark spot in a line of text. That is why they should be widened, not just to fit some classification scheme. Three should be narrower because it is not closed. The difference in horizontals of 5 and seven is a question of how much is going on in the rest of the figure. You wouldn't want a wide horizontal on the 5 to make it topple as top heavy where as the 7 needs something to hold that very long diagonal.

ChrisL

Great topic, Hrant. I have an additional question to throw in here. Dan and Chris bring up some good points about width issues. I love "old-style" numerals and actually have been tossing around the idea of including them with the typeface I'm working on now. After some thought I asked myself (and now I ask of all of you): aside from style or look alone, what purpose do "old-style" numerals actually serve?

Functionally speaking I can't justify using them in heavy text settings and where issues of width proportion are a big concern. I like the way they look though. :)

Paul, I think your progression mostly makes sense (although I think I'd swap {8, 0} and {6, 9}) except the glaring thing with wanting 2, 7 and 4 to be the widest. Are you sure? As for the zero, considering text fonts are meant to have numerals that don't distract, I can't agree with the perfect-circle form. BTW, that's one great thing about 3/4 and hybrid nums: the zero can't get confused with an "O"/"o".

Dan: So you're saying that even OS nums should be monospaced? Or that they're entirely pointless?

> the 7 needs something to hold that very long diagonal.

But then you're magnifying the gap.

hhp

No, they are so useful! But I like them monospaced, or almost so. 1 is always a problem, and you've gotta be careful with the zero, so that it doesn't read like a lowercase o.

And what's interesting is that OS figs are actually easier to make monospaced!

hhp

"But then you’re magnifying the gap."

Not as much as you think. that same diagonal allows tight spacing and the proximity makes the neighbors feel closer.

ChrisL

"...what purpose do “old-style” numerals actually serve?"

Terry,
Each style of figures has a purpose. Monospaced are needed for tabular uses; proportional lining are good in text where you want numbers to be prominent (phone directories, business cards, etc.). So called "oldstyle" is best for running text where you don't want figures to pop and become obtrusive. What you have to think about when designing a typeface is what your users/customers will require and see as a plus in their purchase decision making. Even if you feel comfortable not using oldstyle in your work, your users probably will. Remember, people who actually BUY type are more likely to use it at its fullest and want their options open. This assumes you are talking about a text face. It is less true for display and not needed for signage.

BTW, How is life (and work) in San Francisco?

ChrisL

I am also not a fan of the pure circle zero. It is okay for the "degree" glyph. The circle zero looks like a symbol other than an alphanumeric glyph. The circle is a trap like a bullseye which is hard to escape. It may work OK in very geometric monoline sans fonts. I do feel we have to be careful to differentiate between zero and lowercase o as we do between Zero and uppercase O

ChrisL

1 is a very difficult character to monospace. I've usually seen them with the serif on to get them to work. The zero is tricky because as Chris states there is the problem with the lowercase "o". I wonder if there is a way to compensate for that by making the zero slightly taller.

I personally like the width differences in "old style" characters, but I wonder how the same feel can be kept and still keep them as monospace as possible.

> making the zero slightly taller

That's an idea. Although maybe not "slightly".* Just make it an ascending numeral (hopefully still shorter than cap height though).

* Unless we really want to get medieval on our asses and apply that idea to the entire alphabetics - which might indeed rule. The Greeks used to do it, until Porson threw Modernism at them.

> as monospace as possible

There's little point to this. Tabular use (Dan's point) needs all the nums to be the exactly same width. If you give that up (basically for the same reason that letters aren't all the same width) there's no point approaching it.

hhp

Although I have seen some *relatively* narrow 3's and 5's in historical oldstyle figures before, the Elzevir examples seem particularly extreme to me, and perhaps more so than I am comfortable with.

I do agree that there is room for more width variation than one normally sees in OsFs, without it necessarily being obtrusive.

Cheers,

T

> extreme

Maybe because we're not used to them? Are they more extreme than the "t" and "m" for example? Those two can be made much closer in width than they conventionally are without causing an outroar - but I think most [text] type designers would agree that's a bad idea.

hhp

The 3 and five in the posted Elzevir sample above look squashed to me. I can't see the benefit of such a proportion.

ChrisL

I just added an image up top. It's kind of the opposite
of what this thread is about, but man, is it tight or what.

hhp

Wierd, the 3,4,5 are not dropped but ascend?

ChrisL

It makes the line look heavy as if it were lining figures.

ChrisL

The thing that struck me about the Elzevir sample is that differentiation might be improved if the 3 and 5 were not so similar in width, and/or maybe not on the same alignment. If the 5 were closer to the width of 2.
btw someone put a couple of 1's back in the wrong case in the latest sample.
Tim

> the 3,4,5 are not dropped but ascend?

That's the old-school (19th century) French style, which De Vinne once called "a novelty that failed to catch on". As far as I'm concerned it's worth reviving (or at least thinking about, which too few designers do, even accomplished ones) actually because they fit better with the Cartesian usage of the lowercase; in the conventional alignment scheme you're more likely to end up with a numeral string that seems to sit too low.

> differentiation might be improved if the 3 and 5 were not so similar in width

That's a good point.
And making one ascending and the other descending might be ideal.

> someone put a couple of 1’s back in the wrong case

Good eyes! What's interesting though is that the tabulation was maintained aligned. Maybe all (or a bunch) of the IN's fonts had all their numerals of the exact same width - smart, since wrong-sort situations (more probable with numerals than alphabetics) wouldn't throw a wrench into composition.

hhp

What do you guys think of a tall "2" in an old-style set?

hhp

I think the essence of old-style numerals is that they contain most of their information in the x-height (1-height), so I don't think it would add to its function or form but equally it might not detract either.
Tim

Old style numerals have variable widths, just like regular letters old style numerals were always referred to as letters. Open type fonts support an old style numeral format. Old style numerals, Old style figures the set of arabic numerals that do not align with one another as the lining numerals.
==============================================
Peter
widecircles