Web font embedding returns: Survey!

Last year I posted about free fonts on the web, and briefly mentioned (in the penultimate paragraph) a new CSS-based approach to allow referencing of specific fonts on web servers from HTML web pages. This approach, promoted by CSS co-inventor and Opera CTO Häkon Lie, seems to be gaining momentum, with suport now coming in Safari via WebKit. My personal take on this is two-fold: for the actual end-users, folks designing/producing web pages, it is far too limited because 95%+ of all fonts can’t legally be used in this way; for font vendors it’s scary because it gives users a reason to make retail fonts publicly accessible on Web servers. In reponse to this, Microsoft has now offered to donate their old EOT web font embedding format to the W3C.

One thing that’s been odd is that nobody has really asked typical Web designers/developers what they think of the different solutions for fonts on the Web, and how they’d do things in this brave new world. To try to correct that problem, I have created the following anonymous survey: Click Here to take the survey.

If you’re a font vendor, I have a different survey here just for you.

So, let’s dig into the details, shall we. The CSS “@font-face” solution using the original fonts is certainly simple. Stick a TrueType (or OpenType) .ttf or .otf font on a server, reference it in your CSS, and in a cooperating browser the text will simply get rendered in the specified font. So what’s wrong with that?

  • Fonts are intellectual property, governed by licenses. The licenses of all retail fonts vendors we’ve looked at (about 30 so far) do not allow this sort of usage which is tantamount to free distribution. Although there are many free fonts as well, they represent something like 2%-5% of all fonts.
  • As initially rolled out, the mechanism only works with TrueType fonts. No OpenType CFF, no Type 1 (“PostScript”). However, further discussions with proponents suggest that it might be extended to OpenType CFF (.otf). [Update: although not iniitally promoted as supporting OpenType CFF, the WebKit implementation reportedly works fine with that format as well.]
  • If it’s being done legally, it would only use open source or similarly freely redistributable fonts. But Web designers who are working with an existing organizational identity already have fonts they need to use to maintain that identity online. 99% of those can’t be (legally) used in this fashion. The argument that there are thousands of free fonts so who needs retail fonts ignores the fact that most of the fonts web designers want or even need to use are not free fonts.
  • Of all fonts, probably only 3%-5% could legally be used in this way.
  • There are only the tiniest handful of “free fonts” that come in even a basic set of regular, italic, bold and bold italic fonts.
  • The number of “free fonts” of decent quality is incredibly small. It’s perfectly possible for people to make high quality free fonts, and there are a few out there. But the majority are made by amateurs and are of abysmal quality. Ascender Corp did an analysis of over 4000 free fonts available online, and found massive problems just at a technical level, let alone aesthetic.
  • Based on all the above, I believe that if this new feature is widely used, it will lead to either a major increase in illegal font usage, a major decrease in the quality of Web typography, or most likely, both.

I’ll take a moment to critique a few specific points made by Håkon Lie in the pieces linked at the end of this article:

  • He cited “Goodfish” as an example of a good free font family. Goodfish fails standard tests of its TrueType hinting tables. It doesn’t have even the basic MacRoman or WinANSI single-byte character set. A couple of glyphs lie about their encodings. Glyph weights are inconsistent. As free fonts go, it’s above average, but it’s not usable for body text, and personally I wouldn’t use it for headlines.
  • “Although the legal status of font shapes is uncertain….” Not particularly. Font shapes can be explicitly protected by design patent in the USA (the very first design patent ever issued was for a font), or by copyright in many other countries (e.g. most of Europe). When instantiated in a digital font, the result is protected by copyright as software in both the USA and Europe. These principles have been upheld in court.
  • “…font names are probably covered by copyright law.” Like other product names, font names are normally covered by trademark law, not copyright law.

So what about the .EOT font format? This was originally created by Microsoft, but they are offering to give the format to the W3C as an open format. Because it involves some degree of encryption and doesn’t put entire font files loose on a Web server, font vendors seem to be more okay with it. Perhaps half of all retail fonts could be used legally on Web pages with the EOT format. EOT format fonts can also specify what URLs/domains the fonts can be used with. End users would still need to check their licensing terms, but the process of creating EOT files also checks font embedding bits, meaning that some fonts simply can’t be used with it (because their vendors don’t want their fonts used in such ways).

The funny thing about EOT is that Microsoft actually introduced it many years ago. The problem was, it was a proprietary Microsoft format, so it was only supported in Internet Explorer (and not even in the last Mac version of IE). You needed to run a custom Microsoft tool called WEFT (“Web Embedding Font Tool,” only on Windows, of course) to make EOT files from fonts. EOT was in competition with another equally proprietary format, Bitstream’s TrueDoc with its .pfr files. Neither one was well supported by common HTML/Web development/design tools, and neither became particularly popular.

Of course, an open spec for EOT and an added decade of pent-up demand might make a substantial difference. It remains to be seen how all this will play out at the W3C.

More links:

21 Responses

  1. Chris Fynn says:

    Hi David [David? Maybe you're thinking of my boss, David Lemon. I'm Thomas. - T]Some comments on the Ascender study of freeware / shareware fonts which your blog refers to…1. “TrueType hinting tables – 8.9% failed (404 TrueType fonts had improper/incomplete tables*)””This test checks for the presence of ‘fpgm’, ‘prep’, and ‘cvt’ tables. If all three tables are present the font passes, if any or all are missing the font fails this test. The consequence of a failure is that the font will be flagged as having errors in FontBook under Mac OS X 10.4.”I suspect most of these fonts were created long before FontBook on Mac OS X 10.4came out. To pass this even unhinted fonts need these tables even if they contain no useful data.The statement “Fonts that have hinting information will have better screen quality in Windows than a font with no hinting information.” is imho not always true – With TrueType fonts bad hinting instructions or poor quality “auto hinting” may be worse than no hinting at all. I’ve noticed the on-the-fly auto hinting in FreeType often renders even many commercial fonts better thean when the hinting instructions actually in the font are applied .2a “Code Page 1252 character set – 80.8% failed (3696 fonts missing one or more characters)”2b “Mac Roman character set – 95.9% failed (4385 fonts missing one or more characters)”Without looking at the details of which particular characters are missing these figures are not very significant.If the missing characters are not used or very rarely used on web pages how significant is there abscence?. I’m thinking about things like mu (B5) cedilla (B8) in the “Windows ANSI” 1253 code page, “approxequal” (C5) and Delta (C6) in “Mac Roman”.For English language only web sites in most cases you could drop many other non ASCII characters in these codepages. (This is just what sub-setting in embedded fonts does.)Thinking beyond these two code pages there are of course examples of good quality free fonts like Gentium which have far better character coverage than 95% of commercial fonts. [Gentium is great! Once they extend it to a basic four-member family, it will be even more useful. - T]Also why do they limit this to Code Page 1252 and Mac Roman character sets? Wouldn’t it have been more realistic to to do a frequency study of characters actually used on web pages and compare the coverage against such data?For a few non-Latin scripts freeware fonts may be the only available choice – or better than any available commercial font.4. “Trademark string – 1.7% failed (78 fonts missing a trademark string”If the font name or foundry name is not a registered trademark why should the Trademark string field contain any data?5. “Copyright string – 68.9% failed (3152 fonts missing a copyright string)Even if the font is freeware imo something in the font should indicate this. Popular free software licences used for require this.6. “Embedding restriction – 30.3% failed (1386 fonts set to “Restricted” or improper fsType)”My guess is a far larger percentage of commercial fonts would be set to “Restricted” [There you'd be wrong. The only two foundries I have heard of doing this on purpose are Emigre and Letterhead. Not even 5% of retail fonts disallow embedding. - T]The bare percentages quoted on the Ascender Web font study page do not seem very significant in themselves – and Ascender doesn’t state the percentage of commercial fonts that “fail” the same tests for comparison. [I agree completely that it would be useful to see what the stats are for retail fonts. But they would vary quite a bit by vendor.... - T]I don’t see this “study” as being all that relevant to the main discussion of web fonts.Even if only 5% of free fonts are “good quality” web developers who don’t want pay for commercial fonts would still have quite a choice. 5% of 4000 is 200. If only half of those are aestheticly pleasing that is still much better than being limited to the default fonts available on client systems. [Personally, I don't believe there are even 200 truly free fonts that are at the same quality level as Adobe's. This will probably get me called a font chauvinist or something.... - T]Perhaps what is needed for web fonts is a W3C font validator similar to what they have for validating HTML, XHTML, CSS etc. [I think that's a bit off in left field. The W3C doesn't seem particularly qualified (or interested) to validate fonts. Plus, as the body of your post kind of implies, validation is really only relevant against some particular purpose, even more specific than "for use on Web pages." - T]- Chris

  2. CFF fonts work just fine in WebKit. See my working example: http://www.typophile.com/node/38534%5BExcellent! I’ll update that point. Thanks! – T]

  3. Kevin N. says:

    Photos, sounds, music, text based content – all of these are copyrighted, yet still available on the web – distributed in easy to copy formats. It’s about time for typefaces to joint the fray. :-)Besides, this should be seen as an opportunity for typeface designers to sell more fonts, not less. Even if the MySpace crowd manages to figure out how to copy them – they weren’t going to buy them anyway – but there are plenty of professionals, that would happily shell out some extra licensing money, if it means they don’t have to make graphic text headers any more (bleh!).That said, my preference is to be able to convert (and subset) fonts into some hard to copy format – like is currently done in swf, pdf, dcr, svg etc. (for example, Mozilla et al, could enable the use of svg glyphs in non-svg documents, like xhtml).

  4. Hi Thomas!I’ve been following your blog for a long time, thanks for the consistently high quality :-)[Thanks for the kind words. Sorry that we disagree on so many aspects of this issue. - T]The web-embedding issue is very dear to my heart, because I think it provides a significant differentiator for the free software font movement.As a design business, I can’t take the risk of being sued for usage that is restricted by contract law (EULAs) or for redistribution that is restricted by copyright law. Therefore, legally redistributable fonts are essential. However, most of the 4000 zero price fonts sampled are not redistributable legally; they were initially distributed by their developer without a fee, but without a copyright license that permits redistribution. Indeed, as you say, many do not even have a copyright notice clearly stating who the copyright holder is.Instead of Digital Restrictions Management (DRM), I think that standards should be forged for ways to inform users of the licensing situations of their fonts. This would work for both the software-freedom and proprietary software communities. That way, both communities can be clear about what they are doing, and users can make a more informed choice about which fonts to use.["Restrictions" - cute. But using that kind of language betrays that you're taking a strong side in the discussion (kind of like calling people "pro-death" and "anti-choice" in the abortion debate). Anyway, I'm not against enabling original fonts on Web sites - but I don't see any reason not to also allow folks to protect the versions of fonts that are used on Web pages. Note that we're not talking about DRM for the fonts that end users are paying money to license, just (lightly) protecting the fonts downstream of that, for the derivative copies the licensed user puts on their Web page. This is not DRM in the same sense that has been used in the music or video industries. - T].EOT brings needless complexity and DRM to the table, and is not in the interests of the free software font movement.[It's in the interests of the overwhelming majority of people who want to use fonts in Web authoring, the people who read Web pages, and at least most of the people who design the fonts. If there's also an alternative using original fonts, it doesn't hurt the free software font movement in the least. Is it worth helping your cause if it it hurts the interests of everybody else involved, including both categories of end users? - T]These “DRM bits” are – like all forms of DRM – unconscionable restrictions and are highly undesirable. If the “original fonts on Web servers” approach was modified to only work for fonts in the same domain, so designers can’t point to a font on somebody else’s server, then this will TOTALLY DESTROY the http://www.openfontlibrary.org project – that exists *today*, with several dozen fonts suitable for this kind of linking, and future projects like it.[That seems like a huge exaggeration. If such a change was made in the "original fonts" proposal, anybody who wanted to could simply place a copy of those fonts on their server. Large hosting services like MySpace or AOL or the various blog hosting places could put a single centralized set of the fonts so that all their users could point at them. What am I missing here? - T]If the DRM in the format [EOT] is a mandatory part, it could perhaps prohibit GNU GPL fonts from being converted into .EOT format, although the SIL OFL should be okay. [Fine, so use those as original fonts instead, no? - T] That’s a worry too, since many existing free software fonts (such as the high quality ones from Red Hat, that provide support for many scripts) are GNU GPL. I’ll ask around about this :-)Also, I’m sad to see you use the confusing term “intellectual property.” This term is designed to distort and confuse very different laws, and has no place in a proper understanding of the laws or their purpose. Basically, some works are copyrighted, and some ideas are patented, and some signs are trademarked – but nothing is ever the “intellectual property” of anyone. It is a real shame when any authoritative person like yourself, or institutions like your employer, buy into the confusion created by this phrase, but the widespread use of the term does not make it valid.Best regards,Dave[Well, society, dictionaries, and folks who work with the legal system all seem to think the term "intellectual property" is useful. I do, too. My sense is that you don't like it because it includes the word "property" which has certain specific meanings that you don't believe attach to "intellectual property." Even if I were to grant that viewpoint, there's plenty of precedent in English for word combinations where the adjective changes the meaning of the noun to the point where the set of things described by the new term has zero overlap with the set of things described by the noun alone. - T]

  5. I’m missing one thing here. Maybe clients want fonts as they are in their printed matter. Maybe designers want to embed fonts. But what do end users, the web site visitors, want? That may depend on your audience, but it is a factor that should be part of the discussion.At least some current browsers allow the user to state not to download fonts (regardless of technology used). There’s a good reason for that: downloading a font (even a partial font) adds greatly to the weight of a web page, and the font the designer had in mind may not be the most legible anyway. Being able to choose not to download a font (just like you can choose not to display images) is therefore an accessibility matter, both for download speed and for legibility on-screen (and it would be nice if you could specify that in a user stylesheet, and not just browser configuration!).[Sure, that's a good thing. As an end user, I want that control too - some designers are going to use execrable fonts and I'm going to be happy to override those decisions. Size is not such a big issue, with the possible exception of users on dial-up connections. If the Web designer uses subsetting and a small handful of fonts, the entire set will typically be under 100K. - T]For those reasons, as a user, I have set my browsers to never download fonts. And as a designer I never use them – and won’t. The web is not paper.[Sure, it's not paper. But web designers and the folks who employ them want more control over the appearance of their pages, and this would really help them. - T]

  6. Steve Rindsberg says:

    One thing that concerns me about downloadable fonts is the potential for system problems; one of the oddest and hardest-to-diagnose problems I ever ran across was the result of what was apparently a corrupt font embedded in an document. I’d be concerned that the same thing might occur as the result of fonts supplied by a web page. This was quite a few years ago, and perhaps operating systems are better able to withstand this sort of thing now. All the same, I’ll let someone else dance on the bleeding edge for a while. For now, if my browser doesn’t allow downloadable fonts to be disabled, I’ll switch to one that does. Once bitten, twice curmudgeon, or something like that.[Yes, fonts on Web pages could be an issue for system stability. That being said, both Mac OS and Windows have improved in their ability to withstand bad fonts over the years, so it's not something I'm very worried about as an end user. - T]

  7. Whether it’s an EOT approach or otherwise, there’s a lot of talk of making technology fit around existing font licenses. I understand this being a practical approach, but it does seem to shout out that the font vendors should be creating a new license or licenses that reflects what designer want to actually to with the fonts, ie. have them embeddable in web pages or not.Considering that the font vendors are supposedly so concerned about the advent of widely supported web font embedding, they seem to be doing next to nothing to proactively engage in the discussion (your good self excepted). Web designers are font vendors’ customers too – I would have thought it would be a good idea to listen and act upon their customers’ desires.You also state that “all retail fonts vendors [...] do not allow this sort of usage”, does not Ascender’s server license allow precisely this?”Application Server Font License [...] allows the fonts to be installed onto a server for use with an application that is accessed by an unlimited number of users [...] i.e. for Internet-based applications”[I couldn't agree with you more on most of this. In reaction to these developments, font vendors will need to modify their licenses to make it really clear what is and is not allowed by their licenses in this regard. In some cases it really isn't clear today, which is why in my own corresponding survey for font vendors I ask the question directly, about what is and isn't allowed in their current license terms. It's possible the person who looked over the licenses made a mistake regarding Ascender's license terms. I'll certainly be publishing info on both surveys once I get enough responses. - T]

  8. Chris Fynn says:

    T: [Personally, I don't believe there are even 200 truly free fonts that are at the same quality level as Adobe's. This will probably get me called a font chauvinist or something.... - T]- Perhaps not, but even if there are only twenty or so that means there are as many as the the number of “standard fonts” which most web sites currently rely on which are availible to those who demand that kind of quality and wish or need to use only free & open source fonts.[Agreed. There's no question that enabling original fonts on Web servers would be a significant improvement over the status quo. My main questions are (1) whether enabling that fraction of the world of fonts is what Web designers want and need, and (2) whether enabling that without also offering an alternative that is more IP-protective would simply be a huge incentive to use fonts illegally. - T][I'm sure you'll agree that there are also plenty of commercial font offerings which don't come up to the same quality level as Adobe fonts.]A number of good free and open source fonts were developed by professional type designers or foundries and then made freely available.In some of these cases font development was funded by an NGO or a national government and the fonts made publicly available. In other cases foundries themselves have mad a few of their fonts freely available.So long as EOT is an open specification and available in CSS alongside embedding unmodified fonts I don’t think there is an issue. If it is proposed that EOT should become the only way W3C sanctions font embedding there maybe a concern. – (That would be a something like saying only GIF should be supported for graphics – no PNG or JPG.)[I haven't seen anybody suggest that EOT be the only mechanism, just an additional option besides having original fonts on Web pages. - T}Font vendors could of course choose to license that their fonts permiting embedding only through EOT.Personally I think the Ascender survey and the whole issue of free and open source fonts and their quality or lack of quality is a red herring in this particular discussion.The real issue is whether or not W3C should endorse EOT as a means of embedding fonts which might provide some level of protection to font creators and vendors - thus providing a permissable means for the embedding of some commercial fonts.[I think the quality issue for free fonts (or the lack of variety of high quality free fonts) is part of the reason why web designers want to be able to use commercial fonts, and the W3C should endorse a mechanism that is acceptable to commercial font vendors. But maybe you're right that it's a red herring - regardless of reasons, the main point is that web designers strongly *want* to be able to use such fonts, and the W3C should be responding to that desire. - T]If web developers / publishers embed original fonts where this is not allowed by the license that is no different from those who make available copyrighted text, graphics, or music through their site without permission.- Chris[P.S. sorry about the name mix-up in my earlier comment. The first notice I saw of this blog was in a post from David, hence the mix-up.]

  9. Ben says:

    What is the proposal for a revived EOT? If it is to be encrypted*, I can imagine that this plays into the hands of Apple and Opera – browser authors whose source code bases are not [entirely] open or free, while leaving Mozilla derivatives in the cold. That seems a shame, given the quality of Firefox and the fact that users are very fond of it, and they’ll never know or care what they are missing if they have no implementation of embedded fonts, because the page’s typography will degrade gracefully just as it does now. For Latin type that will weaken the argument for adoption; for non-Latin type it might prevent a lot of people from being able to view text in a good-quality face.*seemingly conflicting with an ‘open spec’ – or at least with an open spec that includes everything required to make it work[Microsoft has submitted the EOT spec to the W3C, including the encryption. So although it is encrypted, the encryption spec would be public. The only part of the current EOT spec that is not included is Monotype Imaging's Microtype compression. How to handle that is still under discussion, as Monotype has expressed a willingness to grant a royalty-free license, but it's not clear whether their terms would be broad enough for all. - T]

  10. BTW, Hoefler & Frere-Jones already covered font-face in their EULA:http://www.typography.com/ask/topics.php?topicID=10“The emerging @font-face tag within Cascading Style Sheets (CSS) will hopefully lead to a secure technology that allows fonts to be used in web pages. But at this time, no such security measures exist, so the use of this tag with our fonts constitutes the illegal distribution of the font software. This type of use is therefore specifically prohibited under our End-User License Agreement.”[Nice to see that they're explicit about it. Give things another year or two and the rest of the font vendors/foundries may catch up! - T]

  11. Chris Fynn says:

    T: … My main questions are (1) whether enabling that fraction of the world of fonts is what Web designers want and needPretty obviously most commercial Web designers would love to be able to use commecial fonts. (I don’t think you really need a survey to tell you that.)[Personally, I would have agreed. But in the discussions within the W3C's CSS working group, some people argue that it isn't important. - T]T: (2) whether enabling that without also offering an alternative that is more IP-protective would simply be a huge incentive to use fonts illegally.It would of course be a huge temptation – but, unless the fonts were modified in some way, the crime would be eminently detectable by automated means – so I think most commercial and institutional web sites would pay the license for fonts they used.[Agreed - although that assumes the license would permit such usage. The problem would not likely be that the big sites wouldn't pay for using the font, but that the license wouldn't allow it. Plus, anybody else could download the font. - T]I think EOT changes the font name, subsets the font, and has some code in it so that an EOT font only works from a specified site – so no two EOT versions of an original font would be the same. I don’t know – but wouldn’t this make it more difficult to detect illegal use of fonts on a web site? If so, might not this disadvantage outweigh the (limited) security offered by EOT?[EOT doesn't change the font name in the file, but you're otherwise correct. You may be confused by the fact that in Microsoft's current implementation, they change the font name when temporarily installing the font, presumably to avoid name conflicts. But this is not visible to either the author or the end user viewing the Web page. Thanks to Si Daniels at Microsoft for explaining this bit to me. - T]It seems to me that font embedding represents a giant showcase for creators of orginal fonts in general and huge potential market for commercial font vendors in particular. The overal size and value of this market could easily grow larger than all other markets for fonts put together – so of course its important to get things right.[There we agree, for sure. - T]T: I haven’t seen anybody suggest that EOT be the only mechanism, just an additional option besides having original fonts on Web pages….T: Microsoft has submitted the EOT spec to the W3C, including the encryption. So although it is encrypted, the encryption spec would be public.So long as the EOT spec is truly open (not another version of the OOXML story) and EOT is not the only mechanism this is probably a sensible way to move forward on something where there is a real demand.Seems to me embedded symbol or pi fonts would also be a very handy way of embedding small scalable graphics (logos, borders, etc.) in web sites.[Interesting. So you see this as a potential substitute for SVG? - T]T: The only part of the current EOT spec that is not included is Monotype Imaging’s Microtype compression. How to handle that is still under discussion, as Monotype has expressed a willingness to grant a royalty-free license, but it’s not clear whether their terms would be broad enough for all.Font embedding must represent a huge opportunity for Monotype to license their library of fonts to commercial web developers. Wouldn’t the value of that market to them far outweigh whatever income they get from their microtype compression?[Yes, but I'm not sure it's an either/or choice. Among other things, EOT could be adopted without the Microtype compression. - T]If necessary I’m sure someone could come up with a an open source form of font compression which would serve the purpose.[That sounds like a reaonable alternative, if the Microtype licensing story doesn't work out. - T]BTW – where are SVG fonts in all this?[I think that using "real" fonts, or EOT (which turn back into real fonts for rendering), are both more versatile solutions than SVG fonts - T]- Chris

  12. Randy Bruder says:

    The nice alternate to a CSS method is using a flash file with the text to display being passed in. This allows you to put anything in any typeface and score browser compatibility across the charts (and keeps the fonts embedded in a .swf and a bit harder to get at)[Of course, at Adobe we're all for people using SWF. But the W3C is in the business of endorsing open formats that could work more simply with HTML. - T]

  13. Atilla says:

    Hello Thomas.Who is the recipient of the survey? Will the result be open to public? What will happen to all the personal data?Thanks,Atilla[The main purpose is to provide info to the CSS Working Group of the W3C (WorldWide Web Consortium), but I will also share said info on my blog. I'm planning on sharing aggregate data, not singling out individual responses. For the type foundries, I'm requesting names so I can track who's responded, significantly reduce the chances of people other than font foundries/vendors giving bogus responses to influence the results, and give me the chance to contact them if I have questions. I intend to keep the contact info and responses by specific foundries confidential, although I may use that info to make some general comments about, say, opinions of large foundries. I *might* also list who some or all of the companies that participated in the survey were. - T]

  14. Chris says:

    The only point I’d like to make is that whatever we do it would be nice if we did it sooner rather than later. Increasingly the type we’re seeing is on a display not paper. I’m sure there are a large number of folk who already eschew paper in favour of some mobile device.Despite this, we’re still limited to Verdana, Times, etc. Font replacement techniques are nice but still lacking in many ways.For myself the inability to use certain faces on the web has precluded my buying them. I only work on the web. I can’t recall when last I printed something.A perfect example: FF Meta Serif looks loverly. Nothing I’d rather do than invest in that face but before I even think about the required bank heist to buy it I realise, no one else will ever see it but me. So, instead, I try to think of a way to make Verdana and Times look less boring (I’m running out of ways to do that).The first foundry to solve this problem conclusively in league with the W3C wins. They win the moral victory and the number to my credit card.Oh, and if I still have your attention, Foundries, why do you bother making “display” faces? I can’t use them anyway, remember?[I didn't understand that last question of yours, until my colleague Miguel Sousa pointed out that you are probably assuming that "display" means "optimized for on-screen use." However, in fact the term goes back a century or more, and means "intended for use at largish sizes" such as headlines and the like. This is covered well in our page on optical sizes in type design. Of cource, fonts optimized for large sizes in print are particularly bad on-screen. You'd be better off with fonts optimized for very small sizes of print, which we sometimes call "caption" fonts. Also, the reverse is true: fonts optimized for screen usage tend to work well at small sizes in print, too. - T]

  15. We spent the better part of a day debating this at the last CSS3 Working Group meeting and my opinion is that, while generous, Microsoft’s offer to the W3C is not appropriate. In my mind, the font foundries (large and small) and browser vendors should get together and determine a set of requirements for an open format that meets their needs and does not expose developers (or browser vendors) to legal problems. If they all buy into it, it will become the de-facto format for the web and if it is made open (like PNG), there will be no barriers to its usage. The main thing we (as designers/developers) will need is tools to convert our fonts to this format (for free, open source, and homemade fonts) and the ability to purchase a pre-built font file tailored to our URL from a foundry (just like we can currently sell customized PDFs for individuals and companies).[The problem with this idea is that it requires a non-organized group of individuals and companies, who have not managed a coordinated effort in over a century, to get together and not just "agree on something" but create a new font spec by donating people and/or money to the effort. I wish that the type industry were this organized. But why reject the Microsoft offer? There's the issue of the Microtype compression, but other than that...? What's "inappropriate"? - T]

  16. Tom Hooper says:

    I filled out the form, however I noticed that it wasn’t keyboard accessible (I couldn’t use tab/space to change the options), and that it required JavaScript to work…Could you consider using standard check boxes for the sake of accessibility and general ease of use?[Sorry, I just relied on what SurveyMonkey gave me. If somebody has a suggestion for a "better" survey service, I'll consider using that next time I do a survey. But in other respects, SurveyMonkey is pretty awesome, so I'd be looking for the new service to be competitive on other features as well. - T]

  17. Your survey question set seems incomplete, and skewed toward the perspective (and your apparently preferred answer) set forth in this blog post.[You're making two separate statements there. It's certainly true that I came in with assumptions about what Web designers and font vendors likely wanted. I did my best to separate that out from the survey questions, and tried to use more neutral language than I had previously used in discussing these issues. Can you be more specific about how the survey could have been improved? What other questions would you have asked? How would you have changed the wording of existing questions? If you think it's massively flawed, why not do your own? I'd be happy to link to it. SurveyMonkey will let you do up to 100 responses for free, or zillions for $20/month. - T]Why not get on with the @font-face solution? If it succeeds with those fonts that allow such use, isn’t that great for everybody (except those foundries that choose to follow the “music label” strategy)?[Fundamentally, it doesn't solve the problem for Web designers as long as they're willing to stay within the law; the survey results support that contention. At the same time, it's pretty scary for type foundries, if there isn't a method that provides a little more protection for them. - T]Also, Mr. Rutter is correct that current licenses do not constitute the only possible licenses.[Absolutely. That's why in my survey for font vendors I specifically asked about what kind of licensing they would be willing to do in the future, not just what they're doing today. - T},/i>I can't take credit for the idea, but how about a font licensing clearinghouse? Count me as one designer who'd be delighted to buy a "site license" for quality embeddable fonts.[There already are licensing clearinghouses, at least MyFonts.com is essentially that. The problem is getting the licensing terms you want. Precious few retail font vendors are willing to license their fonts to be placed in their original form on Web servers where anybody can download them. Modifying that proposal would help, and something like EOT makes a fair number of font vendors happy (though not all). - T](By the way, please consider fixing the tab order in your comment form. It’s disconcerting to press the tab key, intending the next form field, but to instantly jump to the top of the page.)

  18. ["Can you be more specific about how the survey could have been improved? What other questions would you have asked? How would you have changed the wording of existing questions?"]I rephrased most of the questions in comments to the survey questions, and provided my answers to those questions.[Okay, that will be interesting to look at. For the first few dozen responses I was reading all the comments, but I haven't tried doing so again. I suggest commenting on the blog is a better way to have an actual dialog, but I will be reading all the survey comments as well, and perhaps I'll do a new post including some of the more interesting comments and discussing them. - T]New question:Someone asked about added page weight (a few good fonts can add a few hundred KB, or even a few MB). Just curious–has anyone done any work on server-side subsetting of font files?[Do you mean dynamic subsetting at the time the font is sent down to the UA (client)? Or do you just mean having the actual file on the server be subset at the time the file is put on the server or created? The latter approach has always been part of EOT, and of most "font embedding" approaches such as PDF.Although some system fonts are large, usually due mainly to massive character sets, most fonts are not. Without subsetting, a typical western OpenType font is about 30-50K, while our biggest western text fonts (at about 3000 glyphs due to typographic features and language support) weighs in at about 400K each. Those jumbo fonts will benefit massively from subsetting - even being used for body text I expect you could see 90% weight loss there. East Asian fonts are another matter, and get you into the multi-megabytes. - T]

  19. But why reject the Microsoft offer? There’s the issue of the Microtype compression, but other than that…? What’s “inappropriate”?It’s not the offer I object to, it’s that the W3C is not the right place to create a file format; that sort of thing falls outside the W3C’s purview (though, obviously, they can endorse a tech by showing it used in the specs like they do with PNG).[But the W3C has created many file formats: HTML, CSS, XSLT, etc. I suppose one can argue that this is a different *kind* of file format. - T]

  20. Johannes says:

    Note that the name is Håkon Lie, not “Häkon”. I would presume that he also would find the alternative spelling Haakon acceptable, and maybe also Hakon (as a last resort only). “Häkon”, however, is just plain wrong.[Ack, you're quite right. Not like I didn't know it was a ring accent, either. As a typographer, I'm embarrassed to make such an error. I'll go fix it now. Thanks! - T]

  21. Peter says:

    Re Aaron’s comment:It’s not the offer I object to, it’s that the W3C is not the right place to create a file format…If I understand correctly, W3C isn’t creating a new file format. Rather, the proposal is for W3C to adopt an existing format for use in Web protocols.

Comments are closed.


Thomas Phinney

Adobe type alumnus (1997–2008), now VP at FontLab, also helped create WebINK at Extensis. Lives in Portland (OR), enjoys board games, movies, and loves spicy food.

Microsoft Text/Font Program Manager

Thomas Phinney · October 26, 2007 · Making Type

Web fonts: user survey results

Thomas Phinney · November 14, 2007 · Making Type