In general: Italics followed by upright punctuation often cause collisions or uncomfortable gaps. Since kerning doesn’t work across fonts (and even if it did, who would think of kerning italics against roman?) and the two styles don’t necessarily space well, I usually set the punctuation in italics in these cases. It is a structural hack, but it looks a lot better. This is not a one-size-fits-all approach. Sometimes it is more important to communicate a clear structure.

For the special cases: I don’t know about any rules, but I would at least italicize posessive s. I tend to end the italizing at the next wordspace.

See sections 6.2, 6.4, 6.5, and 6.6 in the Chicago Manual of Style for some rules.

Since kerning doesn’t work across fonts…

Although I am not a big fan of Adobe’s Optical Kerning, this is one situation where it has merit.

What if we have quotation marks or parentheses split, as in the following examples?

He said, "The Elements of Style is the most trusted style manual in the United States."
A style manual (The Elements of Style, and others) can be very helpful.

That's where the Chicago approach poses no problems.
Some related discussion here: http://www.typophile.com/node/58456

.

test

Just to throw another style guide into the mix, Bringhurst says to italicize any punctuation following a word in italics. But he also says that parentheses and brackets should essentially never be italicized, even in the midst of italic text. He traces this to the "roman" forms of parentheses in Renaissance italic faces. You can choose whether to buy that or not.

That's typical poet codswallop.

hhp

Perhaps, and I don't entirely abide by it, but as a rule it does look very handsome.

What he's really doing is treating brackets and parens like math signs are usually treated — as geometrical symbols, and not letters. Whether you italicize them (or design italic symbols, if you're a type designer) is an expression of your opinion on the matter: are they quasi-alphabetic characters, or are they signs that exist separate from the different styles of type that letters embody?

See if "f)" looks handsome...
Parens are mainly for math? I'm not buying that either.

hhp

Excellent point, Hrant. Set in metal, “f)” would present no problem: the actual body of the letters would naturally supply the proper kerning. Sometimes, “new and improved” and “idiotproof” aren’t.

And, yes: the notion that parentheses are mainly for math is wholly unsupportable…

Actually most metal fonts had at least moderate kerns (protrusions outside the body that "float") so if you really wanted to set an Italic "f" and a Roman ")" you'd have to add a good deal of space in between, making for a lousy result. And I would think Bringhurst adores metal typography... You know, "concrete poetry"... ;-)

hhp

TeX had the concept of italic correction for almost 30 years now; it is a property of the glyph specifying how much space is applied after italics when followed by uprights, it can be also triggered manually using special command (actually it is not limited to italics, it can be applied to any glyph), it perplexes me how, after all those years, this concept is not adopted by other systems or font format (a fellow TeXnician was very surprised when he learnt that OpenType does not have italic correction, he just couldn’t understand how people dealt with that).

Maybe because a better name for it would be "spacing destruction"? :-/

But I guess since we can't kern between styles it's a decent way of preventing collision at the hands of amateur typesetters.

hhp

Concerning the problem of spacing and kerning between fonts, here’s an idea suggested by Adam Twardoch in the Build forum some years ago, that I’ve implemented in several fonts—roman parentheses as a Stylistic Set in an italic font (see below, No. 4):

Reminds me of the cool tricks the FSI boys added to Ernestine to make the Armenian component work just right with the "shared" characters (including the parens).

hhp

The eminent works: Guts, Greed, and Glory, The (Un)crowded Span: A Comprehensive Review, (and, for example song titles like: John, John, the Grey Goose is Gone), should be set like the setting implemented here.

In other words, punctuation and parenthesis within italicized titles in Roman text are italicized along with the titles. Punctuation and parenthesis joining or separating or containing italicized titles stay Roman. Where this causes spacing problems visually, tracking should be introduced.

Or: yet another reason Italics should be avoided. :-)

hhp

it perplexes me how, after all those years, this concept is not adopted by other systems or font format (a fellow TeXnician was very surprised when he learnt that OpenType does not have italic correction, he just couldn’t understand how people dealt with that).

Well, actually OpenType does have a way of handling this via the [MATH] table, wherein there is Italics correction. This isn't really implemented the same way as TeX, strangely, but that is another story. There is also the notion of cut-ins/outs for kerning purposes which would can also serve a similar function. Mind you, this wouldn't address the issue talked about here, which is more to do with being able to kern between fonts, which we can't do at the font level. Of course, mathematics typeset using proper Unicode-encoded text would infer a font that has the math italics in the same font as the regular glyphs. The math italic glyphs themselves would usually have differing spacing (and perhaps even glyph shapes) then would their 'normal' italic counterparts—all these constituents together allow for the proper spacing of text within a math context, but doesn't really help if the font doesn't support math layout (or if the typesetting environment doesn't either). There was talk of actually implementing a means of interaction between fonts (in terms of dynamic spacing) within an OpenType Math environment for inline maths (as opposed to display maths), but I don't know that its been implemented anywhere yet.

I’ve built several OpenType math fonts and know about italic correction in MATH table, but its use there is just an abuse (it is (ab)used for positioning sub/superscripts, a left over from TeX which did so to maintain a compact font format in the olden days of tight computing resources). Italic correction in TeX (outside math) is certainly applied between fonts, it is simply an amount of kern that can be applied whenever requested to compensate for negative side bearings of italic glyphs, so in TeX with ({\it italic f\/}) you get f's italic correction applied, as compared to ({\it italic f}):

(this is “plain” TeX syntax, higher level formats like LaTeX automate applying italic correction when switching between italic and non-italic fonts.)

but its use there is just an abuse (it is (ab)used for positioning sub/superscripts, a left over from TeX which did so to maintain a compact font format in the olden days of tight computing resources).

Agreed. Its implementation is a bit odd, and in relation to math kerning and the existence of unique math italic characters, somewhat redundant. Its application to integrals is another issue. Alas, it was pretty much a done deal by the time we got involved in building the first (OpenType) math font.

I don't know who said parentheses are mainly for math, but it wasn't me.

I very much like the stylistic set solution, though, for prose as well as math.

The bottom line is, no amount of idiotproofing is going to replace the efforts of skilled operators. Once upon a time, you actually had to know something about a “chosen profession” in order to profess it, and get paid to do so.