>You are right. The code in Axis does not have this bug, but I noticed that >the deserializer code is extremely lame. switch (source.charAt(0)) {
Hah, no wonder I didn't understand how it worked! Well, at least the Axis implementation is efficient :-) >Any suggestions? I would hate to make the code slower by adding >complicated casing checks. The XML Schema doc I'm reading at http://www.w3.org/TR/xmlschema-2/ says there are exactly four allowed literals: {true, false, 1, 0} I don't know if that's case sensitive or not. The same document says that only {true, false} are canonical. I believe that's all Axis emits. In my research I found some griping about the English-centric nature of this coding, but no resolution. Earlier documents talk about "yes" and "no" as also being valid literals, but I don't see that in the W3C docs. Isn't it amazing how even the *most simple datatype possible* is difficult to implement correctly? [EMAIL PROTECTED] . . . . . . . . http://www.media.mit.edu/~nelson/