[BrianLy]

There were likely considerations that fed in at different levels of the process:

1. Higher-ups who still believed that they could offer a platform that would be more valuable than the internet at large. Hailstorm always felt like an AOL for web services. 2. Lower down in the MS hierarchy various areas were under pressure. They had technologies built on COM and Win32 that had been thrown together quickly as a response to the growth of the web. eg. IIS and COM+. Both had problems in terms of scalability (and security as exploited later) and there is was no way to improve performance without jumping down to C++. Obviously there were a lot of other issues with these tools and a "rewrite" was needed.

Microsoft were really playing catch-up here and "copying" various things from the JVM ecosystem such as garbage collection was required. It is worth reviewing Brian Harry's note re: reference counting. A large number of developers were concerned with the implications of this change but it was for the better (for more people - there is still no perfect solution here). C# appears to the uninformed to be a mere copy of the Java language but Microsoft included features to attract some C++ developers, help the transition, and improve usability over Java. SOAP is an abomination, but it is still better than DCOM which came before.

I think Rotor was one of the most valuable things that Microsoft produced in the early years of .NET. Ted Neward et al wrote a book on it with O'Reilly and an unpublished 2nd version is at http://callvirt.net/blog/post/SSCLI-20-Patch-for-VS-2010.asp.... You may not want to look at Rotor if you contribute to Mono-based projects.