From: gnu@toad.com (John Gilmore)
The code is up for FTP where you-all can get it. I made both compressed and gzip'd versions (gzip gives smaller files than compress, is faster to decompress, but slower to compress).
-rw-rw-r-- 1 gnu cygnus 2571835 Feb 5 16:04 celp.speech.tar.Z -rw-rw-r-- 1 gnu cygnus 2099441 Feb 5 16:09 celp.speech.tar.z
Much of the tar file is samples of compressed and uncompressed speech, (used for testing the code). The actual C code is about 340K uncompressed, and there's also a Fortran version in there.
I have a copy of the actual compression standard, but not online. The Information Liberation Front is welcome to a copy -- maybe I should just leave it on the table at the next meeting and hope someone "anonymously" picks it up and scans it in. It's public domain, so there's no special thrill from liberating it.
It occured to me that some people might not get the significance of all this, so prehaps I ought to amplify. With the ability to compress speech down into the same baud rate as, say, a V.32 modem, all one would have to do to have perfectly secure voice communications is replace your phone with a setup that took in your speech, digitized it, compressed it, encrypted it, and sent it over the modem to the other side where this would be inverted. Fast enough software compression of voice would mean any PC with a DSP card and a V.32 modem could become an unbreakable scrambler. The chief problem is that the DSP needed to do decent compression is very crunchy, and encryption also tends to be crunchy, so there aren't typically enough cycles on your average PC. Of course, were someone to commercially market a board that did all this in hardware... Perry