Will GIFs Be Free in 2003?

Ark42 asks: "Did the Unisys patent on LZW expire back on Dec 10, 2002? Does that mean we can all write GIF software royalty free now? From what I can gather, Unisys only lists patent number 4,558,302 for covering LZW, which was filed on Jun 20, 1983 and issued on Dec 10, 1985. According to this site patents filed after Jun 7, 1995 last 20 years from the file date, and patents on or before then last 17 years from the issue date. That means the LZW patent expired on Dec 10, 2002. Am I missing anything?" A deadline of 2003 was given in this earlier Slashdot article. Assuming .GIFs can't follow in the footsteps of Mickey Mouse, will the popular image format now be "web safe"?

Comments

Expires on 20th June 2003.

[kyz]

US Patent 4,558,302 was filed on the 20th June 1983. Under the laws at that time, it expires exactly 20 years after being filed.

fp.

Re:Expires on 20th June 2003.

[Groote Ka]

The original posting is correct, indeed. What needs to be added is that for patents issued after June (?) 1995, but filed prior to that, you may choose between 20 years after filing or 17 years from grant.

User -kyz is confused by the patent law everywhere else in the world, where a patent lasts up to 20 years from filing.

With the change of patent law in 1995 by the US, they finally decided to harmonise patent terms with the rest of the world.

A small step for a human, but a giant leap for such a conservative country.

June 2003

[\\]

According to ImageMagick's file formats page, the LZW patent expires in June 2003.

Re:Gifs want to be free

[damiam]

IE supports just as much transparency for PNG as it does for GIF. It's just that it was designed for GIF's, which only use one level of transparency, so it can't handle PNG's extra levels. It's quite easy, however, to create a PNG with one-bit transparency, and such a PNG would be superior to a similar GIF in every way.

Even better than pngcrush

[friedegg]

Is PNGOUT by Ken Silverman. Best kept PNG secret out there. It only works on Windows (console), but it almost always lets me squeeze a few more bytes out than PNGCrush.

Re:So what?

[kyz]

bzip and rar are both high-performance archivers. They're not intended to work at any high speed. LZW is a high speed algorithm intended to compress quickly rather than compress well. LZO is also a high-speed algorithm, but it compresses slightly better than LZW, and the decompression speed of LZO is far faster than LZW.

bzip2 (note the 2) uses simply burrows-wheeler block sorting with move to front compression, with huffman as the entropy encoder. It will remain like that forever, and not introduce any more compression algorithms. In fact, bzip (version 1, before bzip2) used arithmetic coding instead of huffman, so it actually produced better compression, but IBM et al have a bunch of patents on arithmetic coding, so bzip2 will never use them until they expire. Block-sorting is a "clever trick" to LZ compression, but it doesn't "scale", i.e. you can't put a better predictive model into it and get better compression, the best you can do is put a better sort algorithm in, and we all know that sorting is pretty much at the limits already. RAR on the other hand uses a whole load of algorithms, including Dmitry Shkarin's PPMII which is a statistical compressor that outperforms pretty much anything (at the cost of being very slow). It also has a range of "multimedia" filters, i.e. special processing for images, audio and executables that make the data easier to compress when the real compression is used. RAR isn't open source. If you want something that stands up to it that is open-source, check out 7-zip. bzip2 is not going to get better any time soon.