[css-fonts] Proposal for a new generic font family "Rounded"

[c933103]

I have noted that in https://www.w3.org/TR/css-fonts-4/#generic-font-families there are now proposal for a generic font family known as ui-rounded. However, it seems a bit strange to have a rounded font family option to resemble rounded systen UI font yet without an independent rounded font option to use other rounded fonts that do not need to be same as the rounded font in system UI. It would be nice if there can be a separate rounded family of font so that there can be more flexibility for users to choose their desired rounded font.

[jfkthame]

I'm unconvinced that adding more CSS "generic" font-family names is a good path forward. The set of possible "generic" names for certain classes of design is pretty open-ended (not to mention highly dependent on writing system), and requests will likely be driven as much by the winds of design fashion as by any principled model or taxonomy of fonts.

(As for the justification that this parallels ui-rounded: ISTM even the ui-* names are a bit problematic, as they may not actually achieve what authors want. Suppose I design a page to match current macOS UI conventions by using ui-serif, ui-rounded, etc on appropriate elements. But then a future revision of the OS decides to change its typographic conventions so that the elements that in Catalina were using a rounded font now appear in a slab serif instead: my page suddenly becomes a misfit. System UI conventions have changed in the past, and it seems highly likely they will change in the future. For an author who wants to match the system UI design, it seems like the old system font keywords (caption, icon, etc) would be a more reliable approach than directly calling for ui-rounded, ui-serif, etc.)

[litherum]

I proposed a set of criteria for adding new generic font families in #4910. Perhaps we should finish determining these criteria before deciding on this issue.

[litherum]

OS decides to change its typographic conventions so that the elements that in Catalina were using a rounded font now appear in a slab serif

I don't think this problem is specific to the ui- generic families. An OS can replace Times with a sans-serif if they feel like it, and the probability an OS would actually do this is about the same probability that an OS would replace ui-rounded with a slab serif. (That is to say, ~0 probability.)

[kidayasuo]

“rounded” will be a valuable addition for Japan. I do not have a hard number but it is most likely be the most popular typeface category after serif and sans-serif (c.f. links to font category list by two largest Japanese type foundries below). It is an important typeface in DTP and e-books. Both macOS/iOS and Windows (through Office) have the typeface.

It exists in wide range of languages such as Latin, Chinese (Yuan Ti 圆体), Arabic, Thai and I believe in many other languages. Most importantly they are in a consistent style across languages, i.e. sans-serif with rounded terminals. It is one of the few modern, or abstract, typefaces like sans-serif that do not carry much cultural dependency.

https://en.morisawa.co.jp/fonts/specimen/ (click on “Classifications”)
https://en.morisawa.co.jp/fonts/specimen/ (click on the red slide bar icon on the right)

[AmeliaBR]

Some things to consider here:

• Would any user agent expose a different font for rounded vs ui-rounded?
• Would they allow one to be user-customized and not the other?
• Are there user agents / platforms that have rounded fonts but not something that fits the definition of a ui-rounded font (because it's not used in the user interface and doesn't match the other UI fonts)?

When the ui-* fonts were introduced, the argument was that there were legacy expectations preventing serif and sans-serfi from being updated to anything other than Times and Arial/Helvetica. With that perspective, the ui-* generics are the “modern” generics, and the other keywords are legacy fallbacks. There is no legacy argument for a rounded. But the ui prefix implies extra constraints and expectations that might not be appropriate if an author just wants a rounded style for big headings or buttons, without trying to mimic the UI at all.

Also, re:

OS decides to change its typographic conventions

the same probability that an OS would replace ui-rounded with a slab serif. (That is to say, ~0 probability.)

I think you've misinterpreted the concern, @litherum. The complaint against the ui fonts is that if their purpose was to mimic the native UI, you could nonetheless end up with a dated design if a future design trend means that rounded fonts are used for headings (instead of sans-serif) and serif fonts for buttons and so on.

But that's separate from the question of whether there is any benefit to a non-ui generic.

[css-meeting-bot]

The CSS Working Group just discussed [css-fonts] Proposal for a new generic font family "Rounded" #4605.

The full IRC log of that discussion

<dael> Topic: [css-fonts] Proposal for a new generic font family "Rounded" #4605
<dael> github: https://github.com//issues/4605
<dael> Rossen_: myles can you summerize?
<dael> myles: Sure.
<dael> myles: We got a bunch of font families, add a new one.
<dael> myles: 2 reasons in the issue. 1) we have UI rounded already. Makes sense to add rounded
<dael> myles: Backing up to talk about what rounded is.
<TabAtkins> Note particularly https://github.com//issues/4605#issuecomment-619318179 which talks about "rounded" having a simple consistent definition, and which is used as a typographic style across many languages.
<dael> myles: It's a typographic style where terminals of letters are circular
<fantasai> http://www.identifont.com/similar?MZ
<dael> myles: 2nd argument is it's common typographic style in Japan. If we've got serif/sanserif for the west this makes sense in Japanese context
<dael> florian: UI rounded not particularly Japanese, right?
<dael> myles: Correct
<TabAtkins> q+
<dael> florian: My feeling is this interacts a bit with a thing the spec says is possible, but not done, which is map generic font families to a set of fonts
<dael> florian: if you say rounded I would expect it in japanese and latin fonts, but what I think woudl happen is one or the other
<dael> myles: Not sure I agree. UI rounded is a UI style font. I don't think type of fonts here are for UI, but for toher purposes
<dael> fantasai: I think sep. florian wants rounded to give rounded font
<AmeliaBR> q?
<florian> q+
<dael> Rossen_: I want to channel a question on the issue, do we expect different fonts between rounded and ui rounded. I'm aware of some fonts for windows cjk that are way optimized for small and htinner to allow better fit for overall UI that's closer between different scripts. THat's where ui and rounded will map differently.
<dael> TabAtkins: last week talked about larger issue about generics.
<dael> TabAtkins: Rounded satisfies any reasonable constraints for generics. SOunds reasonable to me
<Rossen_> ack TabAtkins
<dael> myles: One prop criteria was 2 major OSs have built in rounded fonts. Is that true?
<dael> AmeliaBR: I think at least arial rounded is on windows. Might be MS office pack.
<dael> Rossen_: I'm not sure off the top of my head. Could get back
<Rossen_> ack florian
<jfkthame> q+
<dael> florian: I think this is a good idea. Support it. Might be a generic font family that's meaningful across multiple scripts. THat's why I raised point that spec says you can map to sets of fonts. IN this case you should. Spec I'm fine, hoping people will do what spec allows
<fantasai> +1 to florian
<dael> myles: I understand your point now
<dael> fantasai: I support what florian said
<Rossen_> ack fantasai
<Rossen_> ack j
<dael> jfkthame: I just checked windows standard list and there's nothing rounded there. I guess that must come from office.
<dael> myles: Are installations of office common enough that it should be considered built in?
<dael> florian: Yes
<dael> Rossen_: No comment
<dael> myles: serious about the requirement, though. I think it's legit that 2 OSs should have it built in.
<dael> fantasai: Clarify to say not nec built in, but a common config of OS
<dael> myles: Yes
<dael> florian: If you buy a computer and it's on it it counts
<dael> fantasai: Windows with MS Office installed is very common. Languages will vary by region anyway
<fantasai> s/Languages/Installations/
<fantasai> s/region anyway/region and language also/
<dael> Rossen_: From PoV of addition to generic fonts I'm hearing this makes sense. Some questions about how this will be impl and actual behavior. Not sure we need to define this now.
<dael> Rossen_: Question here is does new generic rounded make sense to add
<dael> myles: If it can't be demo there are 2 common config with rounded fonts I think I would object
<dael> jfkthame: Hesitant adding any generic while we're heistent what generic fonts are for.
<dael> AmeliaBR: Agree with that. Whatever we think about rounded it's worth first discussing syntax and general overall design goals with CSS and what generic keywords mean
<dael> AmeliaBR: No pressing demand to add right now so let's take time and do right
<dbaron> My Windows 10 system does not have a font called "Arial Rounded", if that's what I'm supposed to be looking for...
<TabAtkins> +1 to Jen
<fantasai> https://github.com//issues/4605#issuecomment-619318179
<dael> jensimmons: Web front end dev use generics constantly, usually as a fallback font. Try and use something tightly controlled but generics are used as fallbacks constantly
<Rossen_> Arial Rounded is certainly part of Office fonts
<jensimmons> if "rounded" starts to mean something (if) — Authors will use it, gladly
<dael> fantasai: Draw attention to [missed]-san comment. They have been working on i18n and on jltf. Pointing out rounded is more significant in Japan and East Asia. Need to consider it. Also unlike other generics rounded is a style that exists so makes sense in that way as well
<TabAtkins> https://docs.microsoft.com/en-us/typography/font-list/arial-rounded-mt Confirmation that Arial Rounded is shipped with Office
<dael> fantasai: He mentions both macOS and windows have those
<dael> florian: You may want to look for maru which is Japanese for rounded if you're searching for default rounded fonts
<dael> fantasai: Fonts in other writing systems it's likely to be much more common to find rounded. SHouldn't be biased to only consider if it's common in European installations
<jensimmons> +1 to that — need for international thinking
<dael> Rossen_: Are people convinced we should add this or not right now?
<dael> florian: I think we should add this, but we haven't decided the syntax. We might want to wrap a new family in a function so might want to wait a bit
<dael> myles: Rossen_ is right we don't need to decide on that. We can say we want a way to trigger a rounded font without deciding syntax
<jfkthame> q+
<dael> myles: I'm also willing to accept MS office as a common config so we've passed the critea and typographically it's valuable so I'm on board
<Rossen_> ack fantasai
<Zakim> fantasai, you wanted to mention kida-san's comment
<dael> fantasai: We draft spec and we mark them so we can draft is and say we're not sure if it's compat yet.
<Rossen_> ack jfkthame
<fantasai> s/them/issues/
<fantasai> s/is and/it and/
<dael> jfkthame: If we're going to add a generic based on a font from MS Office that might be in tension with work going on to reduce fingerprintability with fonts shipped by system
<dael> myles: Two ideas on surface in tension but if goal is putting users in buckets and making sure none of small all users with MS Office is a pretty big bucket.
<dael> Rossen_: We're at time for a break.
<dael> fantasai: SHould we resolve?
<dael> AmeliaBR: On the general concept?
<dael> fantasai: Yeah
<dbaron> Windows version * office version might be smaller buckets though
<dael> Rossen_: Prop: Add the ability to have rounded fonts and we decide on syntax later
<dael> fantasai: Placeholder syntax and sep issue
<dael> Rossen_: Prop: Add ability to expose and target rounded fonts. Syntax TBD
<dael> dbaron: Little nervous about fingerprinting. May need to consider versions. office by version is small buckets and office version + windows version is small.
<dael> myles: Font is only from office it's not a matrix, jsut office version
<dael> dbaron: Windows version from other things
<dael> dbaron: So we expose office version which might be a matrix against windows version
<dael> florian: Broader, though. It's maybe all versions before 2008 vs after
<dael> Rossen_: dbaron enough of a concern to hold off on resolution?
<dael> dbaron: Given discussion about font fingerprinting I'd prefer to hold
<dael> dbaron: TO respond to florian fonts have revisions. They're not atomic and unchanging. Could have differences across versions
<dael> florian: Yes. I would expect larger buckets than each version.
<dael> myles: How do we make progress on this?
<dael> AmeliaBR: Gets to issue of general purpose of generic fonts and how to move forward. If we add new generics rounded is logical. Security concerns are part of discussion about do we want more generics
<dael> Rossen_: We're going back into generic fonts topics. I want us to stop here. It's a fair point dbaron is raising to hold off. Sympathize with myles to find a way to progress.
<dael> Rossen_: Let's have additional conversations about if this can be unwedged by what hte font metrics look like and are the buckets small enough to conern? If not we can resolve later.
<dael> Rossen_: Outside the fingerprint issue people are fine
<dael> Rossen_: We won't resolve for now.

[fantasai]

Summary of the CSSWG discussion was that there was broad consensus to add this, but some concerns about the privacy impact due to fingerprinting. So if we can be convinced that this concern is resolvable, we will (presumably) add a new generic for rounded sans-serif fonts.

Wrt naming, @LeaVerou suggested rounded-sans, which has the benefit over rounded of being less likely to trigger some kind of naming clash.

[jfkthame]

Wasn't there also a higher-level question re syntax for new generics: whether to simply add keywords (which always have the potential to clash with actual font-family names -- an increasing problem if we're going to keep extending the set), or to introduce something distinct such as a generic(...) function?

[r12a]

By my reading, this was only an agreement in principle that a 'rounded' generic was useful, and that further discussion was needed on what generics are for and what syntax to use, before finalising the discussion.

[LeaVerou]

Wrt naming, @LeaVerou suggested rounded-sans, which has the benefit over rounded of being less likely to trigger some kind of naming clash.

I did suggest rounded-sans but not exactly in that context. What I was suggesting was to have generics that are subcategories of other generics, e.g. rounded-sans, slab-serif etc. Possibly even a function with keyword traits e.g. generic(sans-serif rounded) or generic(old-style serif) etc, which could fallback to fewer keywords instead of just whatever's next on the stack.