Cart
[Home] [By Thread] [By Date] [Recent Entries]
From: Al Snell <alaric@a...> >Even in my self-describing scheme >that would be 5 bytes per element (one tag byte saying "HERE BE AN >IEEE FLOAT", then 4 bytes of float[0]). I think underlying Al's comments is that, especially in the new world of datatyped XML (i.e. with XML Schemas datatypes) there are new possibilities for smart compressed XML and binary XML. The SGML binary discussions did not really have this possibility, and we old-timers will have to re-evaluate some of our experience. That being said, I remember Steve DeRose had an excellent article on his (patented) technique for indexing documents using his fully qualified naming system (as used in DynaText), which gave all sorts of interesting reasons why text formats can sometimes be much more efficient that binary formats, even for searching. It was a very surprising and eye-opening article, but I don't know where it was: if anyone has a URI for that article, it really should be required reading for XML people. In particular, that article gave lots of nice reasons why "binary is more efficient/small than text" cannot be blanketed: some of the reasons came down to binary formats that merely serialize pointers, if my memory serves me well. Which brings me back to my previous point that defaulting/inheritence savings of the particular XML and the particular binary format may dominate more than the simplistic binary v. text issue. Cheers Rick Jelliffe
|
