> > Note that version 3 cleanly splits out the ICU (Unicode) classes, > > which are Open Sourced separately on IBM's DeveloperWorks website (there > > are a whole bunch of companies and individuals working on it there). > > I don't understand this - AFAICS they're split out in 2.3.1 as well.
Sorry, you're right. I've been working with both lately. It is split out pretty well in 2.3.1 as well. > > There are plug-in adapters for each code conversion framework. Right > > now there's one for iconv, one for Win32 conversions, and one for ICU > > (but ICU itself you get from the IBM Open Source site). > > So I need one of iconv, ICU or the Win32 stuff to build Xerces? Never > heard of anything other than ICU, so I guess I'll stick with that. Yes, you need to pick one of those. Some people want to do their own transcoding, so it's split out now as a plugin. Since ICU is Open Sourced separately, we didn't want to have ICU being developed in two places (i.e. no fork). So, if we want it, we can get it from the ICU website. Xerces-C could snapshot ICU, if it wants to (ICU is, after all, Open Source). But, the question is "should we?". What's your opinion? > How much different is Xerces than XML4C atm? The changes required to get > it to compile (given a working ICU) are fairly trivial, so it might be > an easy task to fold them into the current tree. Or would it be better > to wait until I'm sure the XML4C stuff is all working properly? Xerces-C is identical to XML4C version 3 (which was not released on IBM's Alphaworks. It is similar in many ways to version 2. But, version 3 is faster, and has DOM L2, SAX v2, and is XML Schema enabled. It also will (soon) have full internationalization of all error messages, which it shares with Xerces-J (in other words, you'll get consistent error messages between the 2 parsers, and translation of the messages themselves only needs to happen once). Mike
