 Yeah, I tweaked the title a little bit. Yeah, so The adobe type team has always been about trying to invent the future and of course The future isn't always what you expect Keeps things interesting I've been very fortunate to be at adobe for most of the time that adobe's been around and I'd like to share some thoughts With you about the changes we've been through I'll start with saying a few words a bit in particular about kaizen The word you saw in my first slide Kaizen's a wonderful blend of East and West Confucian philosophy and American business practice. It's about getting everybody on the team to look for ways to make things better faster and more reliable Lots of changes often small help quality emerge Kaizen is what brought Japan out of the destruction that resulted from World War two Those of us who are old enough to remember when made in Japan was a badge of shame It may it cheap and unreliable now because of 60 years of kaizen Japan makes some of the best quality products around Similarly adobe's fonts have never been perfect But by constantly improving them, we've made something pretty good I'm here to talk about those improvements adobe quickly developed a reputation for quality fonts What many people don't realize is that this isn't because we were all that good when we started We were simply better than the alternatives and that's not saying much When adobe shipped its first fonts in the 85 the only alternatives for desktop publishing were bitmap fonts If you use them at actual size on screen, they looked okay But if you scaled them or printed them, they looked like a pile of bricks adobe provided bitmap fonts too, but they were just for laying out on the screen They were linked to outline fonts that could be printed on a post-cred printer at 300 dpi Or if you spent the big bucks 1260 dpi on a post-cred image setter Higher resolutions came long soon If you stop and think for a minute, you'll realize there was no software for editing post-cred outlines back in 1984 When we started adobe had to create its own It didn't happen very quickly Also, there were serious size problems Early post-cred printers had very little room for fonts and the connection for downloading print jobs to a printer was really slow Making the file bigger by adding fonts slowed it down more so adobe needed to make the font files as small as possible One of the first families in the early post-cred printers was courier Really talented engineer named Linda gas who now makes beautiful quilts Programmed courier in a text editor The original adobe or courier was a set of post-cred paths that were stroked to create a visible glyph Stroke a few pixels wide made the regular weight and a heavier stroke made the bolt and the outlines were sloped to make obliques To fill in for italics This allowed one minimal set of post-cred data to behave like four fonts Saved a lot of space Of course, it didn't look so great the other main families in the first post-cred printers were Helvetica in times Rumor has it they were digitized by John Warnock's sons. I Believe the engineers tried to program these in a text editor too and concluded that they really had to create a tool for editing outlines The new tool was still pretty crude when these families were produced and of course there wasn't anybody who had any experience with post-cred outlines The end result actually looked pretty decent at normal sizes And they got pretty lumpy once that really large Here are the original glyphs from times in Helvetica compared to more recent versions You can see the difference mainly in the time serifs and the Helvetica curves are really wonky Many things happened in the next few years first Adobe hired Sumner Stone somebody with real type design experience who hired other Typographically talented people and of course Adobe kept improving our editing tools until we finally switched over to an off-the-shelf editor in the mid 90s Finally in 1989 line of type officially complained to us that our funky representation was damaging their trademark So we got new data from them and completely overhauled these families Hinting was at the heart of post-cript success No previous approach was able to make decent bitmaps on 300 DPI devices on the fly The first release of post-cript had a very simple model for hints there was one baseline and Five top alignment zones the blue values array There was one set of hints for each glyphs and hence could not overlap Even before the first post-print printer shipped it was clear that more hinting information was necessary to get decent results By the time the laser writer plus shipped in 1986 Adobe had added more bottom alignment zones the other blues array and A mechanism to handle overlapping hints Hence still can't actually overlap but things that would overlap can be stored in separate layers Each layer is loaded when the portion of the outline and controls is being drawn in the rasterizer Revising times in Helvetica wasn't the only big change we made in 1989 Apple and Microsoft lit a fire under Adobe by announcing they were going to ship an alternate format called true type Perhaps the biggest advantage of true type at least from our point of view was that the OS would rasterize type right on screen So people could see the actual type. They were working with instead of just using bitmap approximations Adobe was determined to get there first with live type and we did we shipped ATM The Adobe type manager nine months before true type actually arrived shipped ATM letting people use post-print fonts. They already had That type had been built to look good at 300 DPI it up Mac screens were only 72 DPI quarter the resolution. We've been proofing at The lower resolution put a lot of pressure on hints We ended up re hinting most of the fonts in the library Our first library wide revision We also well actually that was our second library wide revision We also discovered that we needed to keep as many of an outlines hints as possible in each layer So the rasterizer could be aware of the other hints while making its adjustments ATM also introduced a higher quality rasterization model for better results at low resolutions We still use the old model at larger resolutions where it's faster and a half pixel difference is less critical And at really high resolutions. We simply use Postscripts graphics rasterizer This can make debugging Rasterization problem kind of interesting Moving to ATM forced Adobe to redo a few fonts that use non-standard approaches. We finally retired the stroked courier in 1990 Making it a normal outline font that would work fine in ATM We also had a couple font families that used two sets of outlines one for image setters with their High resolution and one for laser printers with lower resolutions The fonts included code to detect the printer's resolution and apply the correct set of outlines ATM's better rasterization let us switch to using the high res outline everywhere These are outlines of high and low res glyphs from the two families optima and itc eras I offset the low res optima i so you can see the difference better Following the release of ATM Adobe made good on its promise to publish the type 1 font format specification I learned there's a big difference between an unofficial spec and an official one Getting the details nailed down and cleaned up took nearly a year But not only did this help promote the format it made us clean up our own technical act Adobe knew early on we need to be able to handle more than just latin fonts We needed to stop we started working with morisawa To get their fonts working with postcript printers Postcript was set up to work with one byte encodings. So type 1 fonts couldn't use more than 255 glyphs Of course, japanese needs tens of thousands Adobe engineers crafted a sort of virtual font mechanism that enabled hundreds of type 1 fonts to function as a single japanese font It was a massive collage But it got postcript accepted by the japanese printer manufacturers A few years later, we created the much more flexible CID font format that we still use today for east asian fonts In time, we also built up our own type development team in tokyo We understand things better when we do them ourselves So this lets us stay on top of the technical issues as well as creating cool original designs Of course, we had a problem with 255 glyph fonts on the western side too Many fonts really need a healthy set of ligatures small caps old style figures and other alternates in order to enable decent typography There wasn't nearly enough room for them in a one byte font We defined a set of glyphs beyond the standard fonts and set them up in supplemental fonts. We called expert sets We learned a few things Only serious typographers were willing to buy the expert sets most people really didn't care And expert sets were hard to use You had to switch fonts to get the glyphs Which had the extra drawback of making it impossible to curn between them and the main fonts And because type 1 fonts weren't using unicode the expert characters used the values of normal characters often completely unrelated This screwed up the text stored in a document This is one of several reasons we started looking into defining a next generation font format John Warnock figured out how to do on-the-fly racerization in post script So we were all thinking about other ways to use computer technology to do new things with fonts I think one of the coolest things adobe ever did was on-the-fly interpolation In what we call multiple masters We put compatible designs at two ends of a continuum or four corners of a square or eight corners of a cube Or 16 corners of a hyper cube And let the user pick any point in between to get exactly the interpolation they wanted Inside adobe there were people who saw three very different uses for this technology John Warnock saw that we could use it to approximate fonts that might be missing in a document We still have special purpose multiple master fonts in acrobat that do this today Other people saw the ability to adjust weight and width Either dramatically or as subtle adjustments for things like copy fitting And some of us were most excited about enabling designs optimized for the specific size being used Capability that the metal punch cutters always had but got lost Along the way with phototype We figured out how to modulate the adjustment most of which happens in the smaller sizes To get optically adjusted designs for virtually any size As an example of the kinds of design adjustments made for different sizes Here's a recreation of four a's cut by claude garamont Scaled to the same point size We could interpolate between these but the cool technology was pretty hard to support Every time we added a feature like transitional designs or intermediate masters We had to do a bunch of engineering work in atm to support it We also faced a classic chicken and egg problem Very few people actually use the interpolation features because they weren't supported in the main applications The main applications weren't interested in adding support for the interpolation features because very few people used them Though We tried to learn a lesson from this when it came time to promote open type The type one format was a version 1.0 kind of effort Lots of good ideas, but a number of things not considered adobe was working on acrobat in 1991 And along with a way to emulate missing fonts. We needed a way to embed fonts in pdf Of course, we needed to keep the files small even if lots of fonts were embedded A couple of our best engineers sat down and wrote a 2.0 version of the format called it the compact font format Embodied everything we'd learned in the first 10 years of type one And managed to reduce the font file sizes by nearly half When we signed the open type agreement with microsoft, it was clear that cff was the format to use for post-grip open type fonts Shortly after we shipped our first open type fonts. We also published a public specification We haven't made any type one font since 1998 There are a lot of advantages to open type, but cff is definitely one of them from my point of view Like cff open type was a second generation effort We've been thinking about what was needed for 21st century fonts and it turned out microsoft was thinking similar thoughts We combined apple's sfnt structure with unicode support to make a much more extensible format than the original type one or true type And of course it had the convenience of a single cross-platform file The other key component was layout intelligence The open type model lets the font supply information about the glyphs that the application can use to do better typography by default This is cool stuff for western typography But it's essential for many non-western systems Here again, we were learning from earlier mistakes We saw how our earlier efforts with multiple masters never really went mainstream Not well supported in apps and thus hard to use We also saw how apple's gx line layout technology which tackled a problem similar to open type was rejected by all the main applications Because it took over most of their choices about layout We were determined to not repeat the same mistakes and thankfully We now live in an open type world Even so we've seen lots of room for improvement Some of the features we originally defined have since been deprecated Plenty of others have been clarified and reworked And still others are even now being added by interested parties chiming in This is kaizen at its best When adobe started the computer screen was a crude preview device and nothing more We had hand-tuned bitmap fonts that emulated a handful of sizes and other sizes were scale bitmaps which looked awful ATM was the first time the fonts were really imaged on screen As I mentioned earlier we discovered we had to do hinting quite differently for that kind of use Version two of ATM introduced simple grayscale anti-aliasing Which is a new concept to most people at the time We explained it as font smoothing This effectively increased the resolution reversing some of the impact of low res screens later LCD screens began to replace the crts That had been standard for so many decades The lcd structure with red green and blue elements in each pixel opened the door to even higher resolution anti-aliasing Both microsoft and adobe took that step, but independently Best practices for font tables are moving target And standards evolve We do the best we can in qe, but we definitely still find bugs after fonts have shipped Then there always seems to be another currency character coming the euro the ukrainian rivnia the turkish lira the indian rupee the russian ruble There's something this required and at least some of the fonts And there are the general design improvements i talked about earlier So any font that's been in the adobe type library more than a year or two has been revised at least once many multiple times This is kaizen in practice But the revisions create a new problem People don't think of fonts as software that gets updated Unless it actually comes with other software like their os They get a font and use it for decades They don't get fixes and enhancements For a small foundry that has a close relationship with each of its customers There'd be no big deal, but lots of people using our fonts don't even remember where they got them Web font services have pointed the way to a new answer We're now at a point where it's practical to treat fonts as a service This doesn't just enable everyone to have the latest and greatest version It introduces an entirely different business model that has the potential to get far more people using far more fonts Which I think will ultimately be good for everybody This is something i've been looking forward to for at least 20 years So i'm excited to see the first steps finally coming together Thank you