Fudging extrema

Craig Eliason

Picking up from a Typophile thread about situations where an extremum node can be omitted (e.g. in a slightly swelling side of a serif), I'm seeking more opinions on this question:

If I have a shallow serpentine curve like this, is it better to forgo the inflection and keep the extrema (top), forgo the extrema and keep the inflection (middle), or have nodes at both even if it compromises the curve (bottom)?

John Hudson

My inclination would be to keep the extrema but, in a curve this shallow, ditch the inflection.

James Hultquist-Todd

Wouldn’t it depend on the context? Without knowing where this curve was in the glyph, I’d agree with John’s post.

Deleted Account

I would think so. Two of these in parallel would be the bottom one, as a single horizontal, I'd use the top one, and the middle for a single vertical. It could also depend on the lowest expected pixels per feature, a little known parameter.

Craig Eliason

as a single horizontal, I'd use the top one, and the middle for a single vertical

Do you mean use extrema if the curve is going in the direction shown, and use inflections if the curve were 90° rotated? What's the reasoning?

Deleted Account

I mean use the point structure that makes the most sense for trying to capture and preserve the shape in a size and resolution-independent form. So, and I'm not saying for sure it'd make a difference, except if it were either side of a narrow stroke, seeing the structure in place would be helpful.

Alex Kaczun

I would be inclined to go with the middle. Simple and most elegant.
Especially since you are dealing with a serif curve treatment and not defining stems or alignments zones, I would even ditch the inflection.
My experience has been that sometimes less is often more.
Add vectors only if the overall effect is improved.

John Hudson

If any kind of hinting is going to be applied to the shape, then points at extrema is usually essential. This is true for both PostScript and TrueType hinting. In PostScript you can often get away without the inflection point, but in TrueType you will often want to apply an interpolation instruction to control the rounding position of the inflection if the curve is expected to be rendered at a size and in an evironment at which that matters. So, for instance the mid-section of a horizontal curve like this in a GDI CT environment rendered at a size at which the overall height of the curve is 3 pixels might jump up and down across a range of ppem sizes; this can be controlled if the inflection point is explicit and can be instructed.