Reuven Lerner makes an excellent case against OOXML at OStatic
Unfortunately, yesterday’s adoption of OOXML as an ISO standard fails on all three counts. To begin with, it seems that the OOXML standard was poorly defined, leaving a huge number of ambiguities and undefined terms. That’s not surprising, given the fact that it is 6,000 — yes, six thousand — pages long, a size which makes it nearly impossible to ensure internal consistency. The large size also ensures that it will be difficult to create alternative implementations; would you like to be the programmer charged with checking that a particular program adheres to all 6,000 pages of the standard?
Moreover, parts of the standard require a programmer to deviate from many other, correct standards. For example, 1900 was not a leap year, as is the case with three out of every four “00″ years. (Thus, 1900 was not a leap year, but 2000 was.) Microsoft got this point wrong when they first implemented Excel, and as a result, the OOXML standard requires that implementers make this same error, for the sake of consistency.
There are also serious questions regarding some Microsoft patents that any implementer will need to use. Microsoft has promised that it will not sue OOXML implementers for patent infringement, but the Groklaw site points out that this statement might be meaningless.
