<!-- transliterated Japanese (rōmaji): <i> ok -->
We ate <i lang="ja-latn">unagi</i>…
<!-- Japanese (hiragana): <i> not ok -->
We ate <span lang="ja">うなぎ</span>…

This is in part based on the spec’s “(content whose typical typographic presentation is italicized)”, which I note has been removed since I wrote this (quote updated). You’re definitely right on using lang, but I disagree with using i for all foreign words (including those in non-alphabetic scripts) — it’s not current web usage or traditional print usage, would be a CSS maintenance nightmare, and most importantly is not covered by the definition of <i>. For that use case we already have the lang global attribute.

The print tradition of italicising transliterated words is a hint to the reader that they’re not expected to be familiar with it and that it’s not a typo. For example the Japanese loan-word idoru (アイドル) is from the English word “idol”, but has a subtly different, culturally specific meaning. Without italicising it a reader might just think it was a typo. As Przemek points out, using a language in a different writing system is enough to “offset” something from normal prose. Non-transliterated instances of foreign languages in English prose should not use <i>.

The spec specifically says “idiomatic phrases”, meaning phrases like de facto. I’ve filed a bug (“‘an idiomatic phrase from another language’ doesn’t cover non-idiomatic transliterated foreign words”) to clarify the use for transliterated foreign words and phrases. While I’m not backed up by the spec (at least not yet , I’ve amended the article so it doesn’t appear to apply to all foreign words (which was never my intention).

If i really is semantically used for foreign words, then it shouldn’t matter what the character set is, it’s a foreign word.

So if you are going to justify continuing to use i for foreign words for semantic reasons, then set up your style sheet to not italicize it when it is using a different character set.

But it looks like the retention of i is based on all kinds of different semantic implications, which to me means it really is not semantic at all.

em has a clear semantic meaning. I see e, and I know it means emphasis. I see i, what is the semantic meaning? There are several different possibilities, no? Same with u and b. They are presentation tags, they are not semantic tags.

With respect to the dfn tag, I’m not sure explaining the etymology of a scientific name needs the tag. I’m not sure what value it would add to the content or how it would enrich the content in any media.

I think Oli meant, that some typographers say that if you have a word in different alphabet (as Cyrillic or Greek) then you don’t have to use extra alternation, as they look differently already.

<i class="taxonomy">Anaxyrus</i> is from the Greek <span class="greek" lang="el">άναξ</span> and means sovereign or king.

I also changed your taxonomy <span> to <i>. Of course it’s no problem if you keep using <span> there, and declare .taxonomy {font-style: italic;} in your CSS.

@Alice & Shaun —

It seems to me (as in the case of u) that semantic meanings are being made up to justify the continued presence of these strictly presentation tags.

While you may regard them as strictly presentational (and if so it’s fine to use something more semantic instead), there are semantic uses that some people are using these elements for, as I covered in But do we need <u>?.

@Shaun — so I’m guessing you haven’t taught any blind people about HTML yet then?

That’s exactly the case! HTML5 seeks to be backwards compatible (at least partly) so that the spec doesn’t break existing websites. However, it also tries to fit these old tags into the semantic-tags-only paradigm – the result being utterly ridiculous definitions.

When I teach HTML I don’t use HTML5 definitions. I say b is bold, i is italic, u is underline and s is strikethrough. Seems somehow simpler, yes?


<span class="taxonomy">Anaxyrus</span> is from the Greek <span class="greek">άναξ</span> and means sovereign or king.


My taxonomy class uses blue color with italicized text. My greek class specifies Palatino Linotype followed by DejaVu Serif if Palatino Linotype is not available (Palatino Linotype has the most attractive polytonic Greek IMHO).

It looks like now I can replace those span’s with i’s – just making sure to update the greek class uses normal face and not italic.

The difference is largely academic, but it does give a nice unarticulated non-textual annotation. Wait, maybe I should be using u …

OK – seriously, I think this is silly. It seems to me (as in the case of u) that semantic meanings are being made up to justify the continued presence of these strictly presentation tags.

No offense intended to Ian, I’m just calling it like I see it. I will continue to use span, because visual markup is exactly what I’m specifying and visual markup belongs in the style sheet.