Can't the "lengthens" case be limited in theory to "lengthens by one bit" because we can simply add a flag bit for compressed/not compressed. (Practically speaking, it would probably be one byte, allowing for additional information such as compression method or options.)
BTW, I think your "50%" was perhaps meant as a theoretical example, not the real life case. I think most compression methods, when applied to appropriate (photos for JPEG, run-length for character data, etc.) files (other than those chosen as special cases) will do a lot better than shorten 50% of the files. Also, the pigeonhole principle applies only to lossless compression. It does not apply to lossy compression such as jpeg. Jpeg may in fact shorten every single photo file. Charles -----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of Gerhard Postpischil Sent: Tuesday, July 06, 2010 11:48 AM To: [email protected] Subject: Re: VSAM Max Lrecl? On 7/6/2010 12:14 AM, Paul Gilmartin wrote: > And that doesn't help. Yes, it does. There will still be cases where any compression produces larger output, but it is more likely that when one method fails, another will show improvement. > Just consider PKZIP in its entierty as a > complex compression algorithm with several subroutines. The > formal proof applies to PKZIP regardless of its internal complexity. While your statement sounds plausible, it's not compelling. Neglecting special and degenerate cases, a compression method will shorten 50% of the input, and lengthen the rest. Taking a different compression method, it also will compress 50% if the cases, but not the same 50% (by my definition of "different"). While there will be overlap, the total number of compressible data have now increased to more than 50%. Gerhard Postpischil Bradford, VT ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
