>>27
I hope you're trolling. These "tagged text" formats are fucking shit. The "XML crap inside ZIP files" is even worse, yet like all shit, it attracts developers like crazy. Now both MS Office and the freetard stuff use them, and Adobe wants to make PDF that shit too ("Mars" I think it was called).
It's funny, Excel 2007 has an optional hybrid format which uses a binary file for cell contents in order to avoid being slow as fuck. The autosave-every-10-minutes is done in binary too.
Files are made to be read by machines. If you want to have source text or interpret them, more power to you, but the final distributed stuff should be in efficient, compact binary form. Then you use GZIP *on that* (besides, just throwing GZIP at everything is retarded: XML-style garbage can benefit a lot from preprocessing before compression, improving both speed and size, but the browser people are too busy masturbating over the tag soup to notice).
XML vs binary, a point-by-point comparison:
* Size: binary wins
* Compressed size: binary wins
* Read/write speed: binary wins, by about an order of magnitude, in some cases two (good thing file parsing isn't much of a factor anymore, unless you are trying to do serious stuff)
* Human-readability: XML wins, still not optimal for reading though (just open the fucking file, for fuck's sake)
* Human-writeability: XML kind of wins, yet for complex stuff it's useless anyway (maybe for correcting a typo or moving some element or whatever, but I hope you're not thinking about editing an ODT/ODS file by hand)
* Code size and complexity: I hope you're kidding. Good thing you can use some monstrous, hideously bloated libraries to process it, otherwise XML would see no use
* Extensibility: tie. A properly designed binary format can be just as "extensible" (read: backward- and forward-compatible, for the unenterprised people) as XML. A random example is the old DOC/XLS/PPT format that served Office from version 97 to 2003 and can still be saved by 2007 (yes, you can open them in 97 even if they use newer features)
tl;dr: Enjoy your bullshit kool-aid and wanking about trivialities, your CPU-bound servers, and your bandwidth bills.