For me, one of the most anticipated sessions at this year's PDC is Doug Purdy's talk on object serialization in Whidbey/Indigo.
I have somewhat of a love/hate relationship with the current XmlSerializer. On one hand, it's very nice to have a nice object-oriented way to manipulate XML. On the other, I find myself often times fighting against the serialization framework -- either because I'm using a schema that doesn't map nicely into OO-space, or because there are outright bugs in the 1.0 framework.
Rectifying two type systems that are acknowledged to be at least partially orthogonal is a really big challenge. The current serialization framework is a really great start -- but I'm very keen on seeing what the serialization team has come up with now that the problem domain is better understood.
Personally, I think the XmlSerializer is the singular most innovative feature of the 1.x .NET framework -- not so much for what it accomplishes in the end, but because of the way it does its work. I'm in love with the whole concept of attribute-driven runtime code generation. It's too bad that that the practical application of this technology in the XML Serialization domain is less than complete. However, for a first effort I think that the 1.X serialization framework is profoundly cool -- and I can't wait to see what 2.0 has in store!
