Because someone didn’t pay attention.

In the foundry type days of metal, fonts generally had no kerning, as we understand it.

Therefore, combinations like df and gy were allowed to butt up to one another (not overlap, of course), and created odd holes in the flow of text, but this was not considered an issue, as there were many such holes, especially after cap T, V, W and Y followed by a lower case vowel. The concept of kerning for aesthetics/readability did not exist.

Consequently, there was no onus on type designers to design italic g to avoid this problem.
Hence oldstyle revivals that are faithful to lettershape, but anachronistic because they add copious 21st century kern pairs to smooth text "color", should find some digital way to remediate the problem.

With the change from oldstyle to modern in early 19th century, the single-bowl italic g solved the problem.
With the advent of historicism c.1900, and the re-introduction of many old style typefaces, the problem re-emerged.

Linotype introduced "two letter logotypes" in the mid 20th century, for those problematic cap/lower case sequences.
Linotype also redesigned italic f to have a straight descender with no hook/ball, in several faces.
Monotype produced special ligatures for its large Garamond Italic, for character combinations such as gg, gf, etc.

Today, the options are:

- add positive kerning
- create an OpenType ligature
- create an OpenType alternate form of f, g, and y, for contextual substitution
- and of course, design the glyphs so that the problem doesn't occur even in plain text
- combinations of the above

Both the OpenType features are on by default

For the record, Williams Caslon Text Italic does have positive values for these two kern pairs. One may disagree with how Bill chose to address the situation, but I don’t think it’s accurate to say that “someone didn’t pay attention” in this case.

As Nick points out, these are historically problematic combinations, especially for designs which attempt a certain historical fidelity.

John Downer has managed to design the Iowan Old Style italic g, f, and y glyphs so that no kerning values are needed in these combinations, while still retaining the essential characteristics of his historic models, relatively uncompromised — no mean feat.

I didn’t mean to be rude by saying ›paying no attention‹. Nick took some more time to elaborate on this kerning issue. But reviving a classical type for digital use there would have been several options to get around this, as Nick pointed out. Unless it was on purpose to keep the historical fidelity. But as printing is much more perfect nowadays it is debatable to keep these ›flaws‹. One wouldn’t mind that as much in letterpress i guess.