Not sure this belongs in history of typography, but I didn't know what the exact nature of it would be. Here goes.
Imprint, the 1913 face by Meynell and his associates, has been digitized in three versions. I personally like the URW cut. It's tight, more legible than the other two (MT and Bitstream), and it tones down the Caslonic look of its origin.
However, the URW digitization seems to me very peculiar. Most of the lowercase has unequal x-heights. Some of the 'flavor' glyphs come down below the baseline very unequally. Examples. Here's /a/.

Here's /s/

Two other glyphs, even stranger. The bottom halves of /i/ and /n/.


To make a long story short, I posted this to ask a question: is this an accident of sloppiness or haste by the folks at URW? Or was there some kind of typographic wisdom behind these slip-ups that I just don't get? Some kind of a secret idea that makes their cut better than MT's or BT's Imprint?
In particular, I'd like to ask:
does this /v/ look right to you? Or was it supposed to go down lower beneath the baseline?
Comments
I can't say I've designed overshoots that follow these patterns, but I can say that overshoots are a signature instance of where designers have to trust their eyes rather than calculated strategies that seem like they'd produce consistency but don't.
As for the miscellaneous bad details it’s probably sloppy work cranked out quickly during the rush to digitize old type. You’ll find lots of slop if you inspect early digital fonts.
Honestly, in set text, it looks fine to me. Only on close inspection do you see those oddities. Anyway, I wasn't going to make a big deal out of them until I looked at the other two digitizations, which are very even across the alphabet. Then I thought that maybe the artist at URW had a better eye, so I thought I should ask the experts.
On a broader note, I used to think all the overshoots ought to be at the same height/depth, until I looked at DTL van den Keere, which is very uneven, and yet it works wonderfully. So, I discarded that dogma.
Honestly, in set text, it looks fine to me. Only on close inspection do you see those oddities. Anyway, I wasn't going to make a big deal out of them until I looked at the other two digitizations, which are very even across the alphabet. Then I thought that maybe the artist at URW had a better eye, so I thought I should ask the experts.
On a broader note, I used to think all the overshoots ought to be at the same height/depth, until I looked at DTL van den Keere, which is very uneven, and yet it works wonderfully. So, I discarded that dogma.
But it can still happen to find professional fonts with, for example, (very slightly) inconsistent stem weights. Not when they are applied for some optical compensation, I mean.
Afterwards it's not possible to tell how the designer defined the amount of overshoot. Start with guides for small and large overshoot, or start exactly on the baseline and adapt later in all cases for visual impression. It leads to similar results: rounded glyphs have more overshoot.
The same is true for the top of glyphs relative to the x-height or H-height (the letter \H is more reliable to represent the height). Of course there are more differences between fonts, e. g. \H in comparison to \h.