Points at extremes?

alex scholing's picture

Hi all,

Here’s a question that’s been on my mind for more than a decade and i’ve finally decided to seek some wisdom from the crowd.

It is generally recommended that you place points at the extreme tops, bottoms, lefts and rights of your glyphs. My question is: exactly how important is this? I mean, what happens if you ignore this rule.

I do a lot of rounded type and especially rounded glyphs can get you in a bit of a headache here, because the extreme points of rounded glyphs are not necessaryly important points for the definition of the shape - in contrast to non-rounded glyphs where extreme points are almost always also defining points for the glyph shape.

So far i’ve always followed the extreme point rule eventough it meant polluting my outlines a bit, but in the fonts i’m working on right now i’d like to keep my outlines clean. And so far i havn’t had any problems with printing or displaying these fonts...

Can anyone shed some light?
Thanks, Alex.

msmiths's picture


I'm sure you'll get a variety of answers about this topic, but I thought this article might really be of help regarding the story behind FF DIN Round. The PDF linked at the bottom of the article regarding it's construction is very informative.

I believe one of the main reasons behind constructing the rounds with extremes as you mention, has to do with hinting, which is covered on page 20 of the DIN Round PDF.



alex scholing's picture

thanks mike!

an interesting read also.


dezcom's picture

It seems the link to the "detailed...pdf" does not work anymore?

msmiths's picture

Chris, here's a copy from my hard drive I uploaded to Dropbox.


dberlow's picture

The most important thing in point selection is the clean and accurate capture of your idea. Nothing should overrule that.

"Hinting", though not nothing, is only truly effective in TT, where a format conversion is going to leave points close enough to extrema, and your example, as shown on the left is entirely hint-able. T1 hinting depends on Zones which your example can fall nicely into as well, as shown on the left.

If you want to prove to Yourself! that the left version is best for font scaling, regardless of hinting, follow these simple steps, and look at the illustration below.

In a copy of your font with both digitizing options represented, change the Em to around 50/1000th of it's existing units per, and scale all the glyphs too.

Then, scale the Em back to it's original size, 1000/1000ths, all the glyphs too, and observe:

The top left contour is the clean and accurate capture of the idea at 1000, unscaled. The top right is the capture of the idea at 1000 according to the prevailing suggestion, also unscaled.

The bottom is both options scaled to 50 units per em and back.

As big as the deformation to both shapes appears, the deformation of the bottom right seems a bit more serious, and directly related to the presence of the extrema.

dezcom's picture

Thanks, Mike!

dezcom's picture

Man, That is an Xtremmaly vivid depiction of the problem! Thanks, David!

hrant's picture

David, I want to make sure I understand: you're saying that extrema can
actually cause problems? In what cases exactly? Sorry for being dense.


alex scholing's picture

david, thanks for your elaborate answer!

about hinting: do you prefer ttf over otf for screen display?


frankrolf's picture

Alex – thank you very much for touching this subject. David – thank you for your clear example.

I have been discussing this topic over and over, and I agree with Alex that his original drawing (left) is just more natural, specially when it comes to rounded fonts.

I mean, we all agree that drawing glyphs with proper extreme points facilitates things enormously, and is a useful thing to do.
However, when placing extremes presents more difficulties than advantages, I’d really rather focus on the shape instead of trying to fit extreme points, distorting my outlines.

It feels to me that ‘points at extremes’ has taken the place of a dogma for many, and although 20 years ago printers might have crashed trying to process badly-drawn fonts, it is time to re-think this assumption in many cases.

BTW: even the original Adobe Type 1 Specification is not as strict when it comes to extremes (page 30):

msmiths's picture

So do you guys think the FF Din Round example is a bit extreme? Considering we're moving towards higher and higher resolution devices it looks like hinting will eventually become unnecessary and with that said extremes on rounded forms will be unnecessary as well.

I've been facing this dilemma as I have an upcoming face that will have a rounded version and I've found that depending on the weight and size of the contour it can be really difficult to attach an extreme point without compromising the form even if the compromise is very subtle. At the same time, I PS Hint all our faces and know that I'll be at a loss with attaching a hint to a rounded end that doesn't have an extreme point.

I also have some older faces that I intend to rework, adding extremes and properly hinting them although, again with say a Thin weight, even at 2000 UPM things get really tight when trying to attach points on such small rounded contours.

Any advice on this?

hrant's picture

> it looks like hinting will eventually become unnecessary



John Hudson's picture

David: "Hinting", though not nothing, is only truly effective in TT, where a format conversion is going to leave points close enough to extrema, and your example, as shown on the left is entirely hint-able.

I think that bit about PS-to-TT outline conversion leaving [off-curve] ‘points close enough to extrema’ needs to be highlighted and properly understood, because if the points are not close enough to extrema then you end up with a segment of curve that is beyond control of the hints and may round in unfortunate ways.

Not having nodes at external extrema also means that the glyph bounding box will be inaccurately recorded. Tom Rickner told me off for doing that some years ago, and I've been faithfully noding my extrema ever since.

dberlow's picture

AS >david, thanks for your elaborate answer!

You are welcome, but with all due respect, that was the simple answer. I've deemed it socially irresponsible to only help create elaborate answers in a forum;)

dezcom's picture

"I've deemed it socially irresponsible to only help create elaborate answers in a forum;)"

LOL!!! Quote of the day ;-)

msmiths's picture

So say your creating Gotham Rounded Thin, or Hairline for that matter… your radius is going to be so tight, even at 2000 UPM that attaching a node to the extrema is going to most likely distort the contour… so what's the lesser evil, distorting the contour isn't an option, so it has to be either not pushing the rounded style family weight range too far or not hinting those if you do, right?

I'm rambling, apologies if what I'm saying is a little confusing lol, but you get what I'm saying.

alex scholing's picture

haha david, sorry, i did mean of course: thanks for the insight, if not wisdom. i now feel completely comfortable with sticking to my clean outlines ;-)

alex scholing's picture

oops, one click too many

dberlow's picture

>so it has to be either not pushing the rounded style family weight range too far or not hinting those if you do, right?


William Berkson's picture

I don't know what is best practice, but it is a myth that the Adobe spec always requires points at extrema. The Adobe Type 1 fonts specs say that "It is not necessary to place an endpoint at extremes of very small curves such as at the tips of curved serifs." See p. 29-30.

David, does this relate to the problem you point out?

dezcom's picture

I take it as meaning that details which are not possible to describe under low rez conditions (including serif details) just clutter the hinting scheme so they are best left without points at extremma. Curves that are of a relative size to be described adequately under low rez conditions still need points at extrema, however.

twardoch's picture


if you define an alignment zone or ghost hint for the purposes of equalizing the baseline and the x-height in very small screen sizes (through hinting), and in some glyphs there will be no point there, or the point will not be at the extremum, then the hinting may work not as you'd expect. In your design, hinting will work better if you draw your glyph as you've shown it on the right (with points at extremes).

Points at extremes are primarily relevant for hinting purpose. Other than that, you don't need to worry about them. Especially in italic typefaces, points at extremes in the x-direction (i.e. on the left and on the right) are not really important if you don't use x-direction hinting.

Ray Larabie's picture

Billy Joel once asked a similar question.

msmiths's picture


alex scholing's picture

thanks adam, until now i didn’t know the extremes were about hinting. i thought they were about accurate boundingboxes and spacing.

i think i’m going to try to stick to my clean outlines and try to solve things through hinting. so far i only did .otf because of simplicity but i suppose i will go .ttf now because of the hinting. study!

twardoch's picture

Points at extremes are important for both .otf and .ttf because of the hinting. Basically, if you want to have your baseline and x-height align at small sizes, i.e. you want the overshoots for your round and spiky letters not to look awkward, then you need some points there (and hints) so the rasterizer knows WHICH points to align. Because of subpixel rendering, points at extremes are still important in the y-direction (i.e. on top and bottom of forms) while in the x-direction they are not. For example, for italic letterforms, I would say that they are no longer really mandatory on the left and right edges. But it's still "safer" to put them there until we can reliably find out that not having them doesn't cause some awkward issues in applications that rely on bounding boxes when doing line layout (for example, I suppose Adobe InDesign's "optical margin alignment" does take bounding boxes into account).

dberlow's picture

>i think i’m going to try to stick to my clean outlines

A good plan, the best actually. Once one screws up the shape, nothing helps, while hinting is really pretty flexible if you really understand it.

Syndicate content Syndicate content