[jws]

In general, anywhere you find Huffman coding being done, it could achieve superior compression with Arithmetic Coding but is not because of IBM's patents.

This is essentially every JPEG image on the planet. The standard provides for arithmetic coding, but no one implements it because of the patent. Wikipedia asserts that arithmetic coding saves about 5% of the files size.

So that is 5% of every flash card sold for a camera wasted and 5% of the bandwidth costs for images on the internet wasted.

If GATT hadn't tacked on three more years in 1995 these would have expired by now, but it looks like 2012 or 2013 is the date now.

The MPEG series of standards probably could also have benefited similarly from Arithmetic Coding.

[astrange]

> The standard provides for arithmetic coding, but no one implements it because of the patent.

That is, there is an implementation in the libjpeg source tree, but it's disabled. I thought the patent had actually expired now, though.

Also, H.264's arithmetic coding mode saves ~20% bitrate; Vorbis or AAC could presumably benefit too, but neither of them use an arithmetic coder, even though AAC should have no patent problems. Maybe there's some technical problem I don't know about.

[eru]

And isn't MPEG (or at least mp3) covered in patents, too?

[jws]

There are about 1000 patents covering MPEG. Just the MPEG2 list runs 36 pages.

But maybe YouTube could be saving 5% of their bandwidth bills if arithmetic coding weren't patented.

[eru]

Or at least if the patents weren't enforced.