ClearType Patents, FreeType and the Unix Desktop: an explanation
David Turner (david at freetype dot org) 2007-06-01

Now that the first Linux distribution has removed LCD-optimized text rendering from their default build, it's time to explain a few things as well as clarify the situation.

I. The ClearType patents

I.1. Meet the patents:

Microsoft lists no less than 10 patents on their ClearType licensings page. In case you're interested, the relevant US patents are:

6,219,025 6,239,783 6,307,566
6,225,973 6,243,070 6,393,145
6,421,054 6,282,327 6,624,828

Essentially, these patents cover several different things, grossly sub-divided into:

Also, it may be possible that Microsoft acquired other patents related to the field in the previous years. Who knows

I.2. Patent claims, prior art and validity:

Steve Gibson claims that the technique used by ClearType is a reinvention of a 20-years old thing used on the Apple II. His exact words are "Thus, Microsoft's 'ClearType' application of sub-pixel text rendering does not represent the dramatic breakthrough that they claim and it can not be the valid subject for intellectual property acquisition". (emphasis mine).

Unfortunately, Mr. Gibson doesn't understand patent law well. Under the current US regime, any minor improvement to a previous technique can be considered an "invention" and "protected" by a patent under the right circumstances (e.g. if it's not totally trivial), If we look at the first patent, we see that the Apple II Wozniak patent covering this machine's display technique is listed first in the patents' citations. This shows that both Microsoft and the patent examiner who granted the patents were aware of this "prior art".

I'm not trying to defend Microsoft here, just want to avoid feeding false hopes to people who would want to see the patent revoked.

Another popular view is that these patents are too general to be enforceable. Well, to be fair, some of the claims in these patents do indeed use rather vague descriptive terms (even for a patent lawyer or an "expert in the field"). this is absolutely not surprising, it's a direct consequence of how the patent game works.

I'm not going to cover this in details (there are many interesting pages on the subject), but will say that even if you invalidate a single patent claim (e.g. with prior art), that doesn't mean the whole patent is busted. Any other independent and dependent claim can still be enforced.

Some of these patents have up to 40 claims. Invalidating them is going to need serious prior-art. Even if there are also strong chances to invalidate the most general claims in there. For example, many of the cheap LCD screens on digital cameras have used a screen where each pixel is either Red, Green or Blue, with colour images directly mapped to them; they've been doing it for years, even when displaying text/menus, and this corresponds exactly to what the most general claims cover. If we can find a proof that the technique was deployed before the patent's filing date, we could have valid prior art to bust these.

I.3. Possible Work-arounds:

People have proposed alternatives to the ClearType method. A very good example is the SubLCD technique, which uses a different way to use the sub-pixels than Microsoft's ClearType implementation. Its author even says it doesn't infringe on the ClearType patents.

Unfortunately, I, as the FreeType author, do not share his enthusiasm. the reason is precisely the very vague patent claims described previously. To me, there is a non-negligible (even if small) chance, that these claim also cover the SubLCD technique. The situation would probably be different if we could invalidate the broader patent claims, but this is not the case currently.

II. FreeType and the Unix desktop

II.1. Does FreeType implements any of the patented techniques ?:

Technically, no. The patents cover the whole process of generating and displaying sub-pixel images. Since the font engine doesn't do the display part, it cannot infringe. Apart from that, FreeType has provided the capability of converting vector shapes into un-filtered sub-pixel images for a long time, but this feature isn't used at all in a typical Unix desktop (see next section).

When releasing FreeType 2.3.0, I've made some drastic changes to the font engine to clarify the situation:

this means that, with a default FreeType build, even if you use its LCD-specific APIs, it's no longer possible to generate a patent-infringing sub-pixel image: the result will always look like "normal" gray anti-aliasing. You can override this limitation by switching a single define in a configuration file though, but you should do that at your own risk.

And the reason I did it is because the font engine is used in countless embedded products, like cell-phones, PDAs, set-top-boxes and even video-games. Even though the software is provided "AS IS" with no warranty, I prefer to avoid putting its users in legal jeopardy when they download my released code. For Linux distributions, things are a bit different because packagers routinely patch upstream packages, but at least that's not my problem or my responsability.

II.2. Impact on Unix desktops:

The latest changes in 2.3.0 do not impact a typical Unix desktop. Because at the moment all LCD-optimized rendering is normally implemented by two libraries: Xft and Cairo. And both do all their LCD-specific processing using FreeType's "normal" (non-LCD-specific) APIs. their implementations differ but produce identical results.

A common misunderstanding is that OpenSUSE removed LCD-optimized rendering because of the changes in FreeType. This is not directly true. the real reason is that they patched Xft and Cairo to either remove their LCD-specific code, or use the newest FreeType APIs instead. In all cases, I believe that this was the right thing to do, and I urge other Linux distributions to do the same. Oh, and I don't think this has anything to do with any Microsoft patent deal.

Conclusion

To make it short: OpenSUSE and Fedora did the right thing by removing any patent-infringing code, and I hope other Linux distributions will follow. If you do it for MP3, do it for LCD text.

And it'd be nice if both the upstream Xft and Cairo could use the newest LCD-specific FreeType APIs. I have even provided patches to do just that a loooong time ago; for some reason, these have never been integrated by the respective library maintainers. I hope some day they'll wake up (though my hope is pretty dim). I suspect that Linux distribution packagers are already applying these patches though...