I've been experimenting a bit with @font-face recently so I was intrigued when I heard about this talk. It was booked out within hours of announcement but fortunately (for me) a few folk couldn't make it and the people on the waiting list got to attend.

Jeff has already done a 'live blog' of the talk, and there's a full transcript on Ben's site, so I'll just give a brief and slightly out of order outline and concentrate on the bits I found interesting.

Ben is a typography geek who got into web design so he had a good understanding of both the typographical world and the history of web fonts. This he discussed for the first third of the talk - explaining what a font is, why font design is difficult, why typography is important and why, more than twelve years after Microsoft first brought web fonts into the world with IE4 and EOT, we might finally have a chance at a solution that works for users, designers and the font foundries. One of the interesting aspects he discussed here was ligatures - where two letters, when placed consecutively, are melded into a single symbol. This happens in English fonts for things like 'ff' and 'fi', although the individual letter shapes remain, but the effect is even more marked in Arabic scripts where the combined character looks significantly different to the two (or more) which it represents. Having to deal with all this sort of stuff is why font design is so difficult, why even the 'open source' fonts have mostly been paid for (one way or another) and why font foundries have been generally paranoid about the possibility of all their hard work being stolen as soon as it's uploaded on to the internet.

Although most of the excited noises about web fonts are coming from designers, there are a few reasons why they're important. Firstly, while the font support for English and other languages based on the Roman alphabet is good, there are others where it's not so good and some (relatively popular) languages which have no font support at all in most operating systems (Telugu, a language of the Indian subcontinent, has 74 million speakers but no support in Windows). In this situation many sites have resorted to delivering their textual content as images. Delivering textual content as images is also an issue for accessibility as many designers, desperate to use their fonts of choice, resort to image replacement techniques which, if done badly, can result in poor accessibility.

This leads neatly in to Ben's discussion of the various hacks people have used because of the lack of working, cross browser web fonts. In order from the stupidest to the cleverist:

• Image replacement - Basically, write your text in Photoshop and save it as an image, then put it in your page. An accessibility nightmare, particularly if the designer 'forgets' to provide the alt content.
• Fahrner image replacement - In this technique the normal text is left in the HTML and the 'rendered' font is put in a background image. CSS is then used to move the normal text out of the viewport, leaving only the background image. Less stupid than straight image replacement, but still not perfect - no support for text resizing and you have to generate all your images in advance.
• sIFR - Moving on to the clever hacks, sIFR works by embedding your font into a Flash movie then using Javascript to replace the headings dynamically. This was the first hack to be CMS-friendly - the text to be displayed is a run time parameter, but it introduced the additional requirements of JS and Flash.
• Cufón - Getting to the first of the really clever hacks, Cufón uses a generator to convert your font into a set of JSON encoded outlines and then uses canvas to render it in the page.
• Typeface.js - The cleverest of all, by making use of a server side component with access to the Freetype library, and then applying according to your existing CSS.

While these libraries get the job done, they do have a number of drawbacks - sIFR's use of Flash, Cufón's requirement to specify your fonts outside of CSS, the way it splits the words up into separate elements for each letter, and the general dependence on Javascript. They also have some drawbacks on the typography front, particularly to do with ligatures and letters that change shapes in different contexts - Ben showed us a number of example slides which you can check out on his site. Also, none of the use 'real fonts' - the same things that any other application would recognise as a font file.

We moved on to web fonts and the @font-face rule. Here is what it looks like:

@font-face {
	font-family: 'CantarellRegular';
	src: url('Cantarell-Regular.eot');
	src: local('Cantarell Regular'), 
		url('Cantarell-Regular.ttf') format('truetype');
}

You can then reference your downloaded font like this:

p { font-family: "CantarellRegular";}

All relatively straightforward, this will work in recent versions of all the major browsers - a standards compliant way to render actual fonts of your own choosing, with no need for any scripting or third party plugins, taking full advantage of typesetting capabilities already available within browsers and operating systems. Of course it's not quite that straightforward, there is some server side setup to consider, and then the ideological and technical hurdles that still need to be overcome before this all becomes practical for mainstream sites.

To deal with the idealogical problem first - font foundries are still not playing ball. This is a problem because, unlike music, a font is software rather than data and gets the same protections, also, as should be clear from some of the difficulties described above, good font design is difficult and so professional designers realise it is worth paying for a good font. So, even though the technology and standards are now getting in to place, font foundries are still unwilling to license 'desktop' fonts for distribution with websites.

And it's not like there are no practical problems. First you've got to consider what adding a few fonts to the mix is going to do to your bandwidth requirements. Regular desktop fonts, with just normal, bold, italic and bold/italic options, exceed 100k in size, especially if they include a full unicode character set. A font from a foundry, which is likely to have many more variations (extra heavy, expanded, smallcaps etc.) can be up to 1Gb in size. This gives you another thing to think about - what will the user see while waiting for the font to download? In Firefox they'll see the page rendered in whatever the default font is - this is likely to have a different geometry to the font you're delivering, so the page will have to be reflowed when the font arrives. This causes usability difficulties as elements users have started interacting with suddenly move around. In Safari the user will see nothing until the entire font is downloaded, making it impractical to use a large font for body text. This is the flash of unstyled text problem. You might be able to subset the font to reduce download times, but then you need a font with a permissive license, which leaves you back at the idealogical roadblock outlined above. You might have the idea of a single repository for web fonts to increase the chance users already have the font you want to use, similar to Google's Ajax library (and you wouldn't be the first), or even just to share a font across multiple domains in a portfolio. In this case you've got to set up Cross-Origin Resource Sharing for Firefox.

Although @font-face itself has wide support there are some differences in the details when it comes to defining families of fonts. Firefox seems to do pretty well and Safari/Chrome behave similarly, there are some issues that crop up in Opera and Konqueror, but most of the problems occur in IE. Here is how you declare a font family according to the spec:

@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif.ttf") format("TrueType");
    font-weight: 400;
    font-style: normal;
}
@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif-Italic.ttf") format("TrueType");
    font-weight: 400;
    font-style: italic;
}
@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif-Bold.ttf") format("TrueType");
    font-weight: 700;
    font-style: normal;
}
@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif-BoldItalic.ttf") format("TrueType");
    font-weight: 700;
    font-style: italic;
}

Each different weight and style has it's own definition and associated font file - multiple font files can be listed and the browser should pick the first format it can handle. However, of the above, this is all IE understands:

@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif.ttf");
}
@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif-Italic.ttf");
}
@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif-Bold.ttf");
}
@font-face {
    font-family: "DejaVu Serif";
    src: url("/fonts/DejaVuSerif-BoldItalic.ttf");
}

Internet Explorer ignores all the font-weight and font-style rules so essentially all that happens (leaving aside the issue of file format) is that DejaVu Serif gets redefined repeatedly. A common work around is to define each font style explicitly in it's own font family, but this is hardly ideal.

In the final part of his talk Ben moved on to WOFF (Web Open Font Format) - the new standard which could finally get the font foundries behind web fonts. It was initially worked on by Erik van Blokland and Tel Leming, two guys well respected in the world of type, then worked on by Jonathan Kew and John Daggett of Mozilla, two guys with a lot of respect in the web browser world - so it was able to gain a lot of traction in both areas very quickly. So the future looks good, in the meantime there are a number of startups and other websites looking to exploit the demand for web fonts in TTF and EOT format:

• Typotheque
• Typekit
• Font Squirrel
• Open Font Library

This was a very useful talk on what is a hot topic in the web design world right now, I was particularly interested to learn about some of the history behind web fonts and some of the issues surrounding support of non latin languages, so 5 out of 5.