PNG Group Unconcerned About Apple's Patent

melquiades writes: "A recent story raised concerns that Apple's patent on some forms of alpha compositing was blocking the development of PNG, MNG and SVG. Not so, says Greg Roelofs, a member of the PNG group: 'The PNG group did discuss the Apple patent several weeks ago, and we decided it was completely irrelevant to PNG itself, almost certainly irrelevant to the pnmtopng utility and to PNGs animated extension, MNG, and probably irrelevant to SVG as well.' Here's the article on OS Opinion. So if it's not a big deal, why was there a general call for prior art to overturn Apple's patent? It looks like some PNG developers got worried, but the core team thinks there's no problem. Is this just a case of the right hand not knowing that the left hand is paranoid?" Once bitten, twice shy?

Comments

Leftover Paranoia

[rsimmons]

Its just leftover paranoia from the problem Unisys had with people using the GIF format. The PNG group seems to have a good understanding of the Apple patent, from what I read in the article.

As I understand it...

[Millennium]

It looks to me like the patent covers alpha-masking, not alpha-blending.

As I recall, PNG and SVG both use alpha channels. That is, transparency data is stored as a part of the image's color encoding.

Alpha-masking is different. In this case, an image is defined in the "normal" RGB fashion. Along with it, a second image is stored, which has only alpha information in it, not color. You could say that an alpha channel is inline, whereas an alpha mask is not.

Is the patent ridiculous? Of course it is. But I don't think it would affect PNG and SVG anyway, because they use a different method of storing transparency information.

Re:As I understand it...

[Doomdark]

I thought so too, after first glance at the patent. However, after reading it through second time, I understood what the patent really is about.

What it is is doing 'compontent-sensitive' image blending. Instead of using single alpha mask or channel for all colour components, it uses a full mask image, that is then used for blending images component by component. When mask image is a gray-scale image, this effectively degenerates to 'normal' alpha blending. However, when using non-gray colours, blending is not linear (as with alpha blending) for the components. In RGB, for example, you could take red value from source, green from destination, and blend blue 50-50 from both, and get... um... probably interesting results?

What I would like to know is if this is useful? What kind of effects can be achieved by using non-grayscale mask images? Potentially it might produce interesting effects... But are those just curiosities?

... and no, obviously this doesn't cause any problems for PNG, which 'only' uses embedded alpha channels.

The actual forum post

[marmoset]

The OSOpinion article refers to Greg Roelofs'
forum post. My favorite part of the Roelofs' post is this classic line:

" So I nominate this for the 2001 tempest-in-a-teapot award... But since the /. crowd lives for this stuff, I wouldn't presume to imagine that my comments will have any discernable effect on the (flame)festivities. :-) "

How many posts did that article get the other day? :)

Apple Unconcerned too?

[Anixamander]

A while back Apple posted this to their web site. It seems they are against RAND licensing fees. I assume that would also include their own technology, but I am really not sure how all this fits together.

Why, you ask?

[Lars T.]

From the the article on OS Opinion: (my emphasis)

Earlier this year, the World Wide Web Consortium (W3C) asked interested parties to submit information about patents they own that might potentially infringe upon the development of scalable vector graphics (SVG).

So Apple reported one of their patents that could in the future conflict with the development of SVG. Sombody read that, combined the two evil buzz-words "Apple" and "patent", and paranoia kicked in. He then scanned through the patent, found "alpha channel", and thought his worst fears had come true. So he called up some buddies and started the usual anti "evil/stupid patent" actions.

All in all, simply self induced FUD.

Noninfringement Opinion?

[Compulawyer]

I know this is Open Source, but I generally thought that most Open Source software was better engineered than much of the commercial stuff out there. One of the key components in a feasibility study BEFORE other development work begins is legal feasibility . If you don't believe me, dust off your copies of Pressman's Software Engineering and see what HE says. I'm not just making this up.

Part of a legal feasibility study has to be a patent infringement analysis. Whether you agree kthat software should be patentable or not, the fact it that in the US it IS patentable. If someone else made it first and was granted the exclusive right to make that product, then you will infringe on their rights if you make the product.

There is simply no excuse for not doing this analysis. Yes, there is some up-front cost to a patent attorney. However, with a large enough development group, this cost can easily be absorbed by the group at a nominal level per member. The potential consequence is that every developer can find themselves defending a patent infringement suit. The last estimate of which I am aware placed the cost of litigating an "average" infringement suit at just around $500,000.00. If the case is deemed "exceptional," the infringers could potentially have to pay the patentee's legal costs as well as their own, on top of paying damages for the patentee's lost profits.

Bottom line: all this can be avoided by getting a noninfringement opinion by a reputable and competent patent attorney. This is a basic software engineering step and should be performed at the very beginning of the development process. The attorney will even be able to suggest ways you can still go forward with the project, but do so in a way that does not infringe on someone else's rights.

That's an utterly ridiculous attitude.

[Nindalf]

While we're at it, let's spend all our time and effort preparing to defend PNG against patents on automated shoe manufacturing and quack magnetic therapy devices. After all, "only a court can decide" whether PNG software infringes on those, too.

The patent itself specifically contrasts what it covers with the use of an alpha channel, such as is used by PNG. It is obviously not a threat to anyone who bothers to read the patent.

Aside from that, PNG was designed to avoid patent issues by people who knew what they were doing, in an drawn-out open process where anyone could have pointed out weaknesses. It's based on well-established techniques with a long history of prior art. Your first assumption should be that any patent either doesn't affect PNG, or is so blatantly invalid that nobody would ever dare take it to court, not that there's trouble ahead because a one-line description of a patent sounds vaguely like something PNG does.

You can't live your life shrieking in terror every time someone whispers "patent."