Monday, November 11, 2002

I have just been given a go-ahead (Visual C++.NET only so I won't talk about the rest) since a lot of this has gone public already. The Everett Visual C++.NET features, I feel, are the biggest changes to VC++ since 5.0 and have made me extremely happy. The features break into 4 main areas: ANSI/ISO Compliance Forms Designer for Managed C++ Security Features Optimizer Improvements particuarly in floating point The ANSI/ISO compliance is particularly exciting as Visual C++ finally does Partial Template Specialization, Partial Ordering, Member Template Definitions, and pretty much everything in the standard. Public statements have said 98% compliant. I have not been able to throw any Standard C++ at it that it couldn't compile and handle correctly yet. Very heavy users of advanced Standard C++ features, particuarly templates like Loki, Boost and Blitz now compile 100%! Also, in the unmanaged arena, there have been significant additions for security, paticuarly in the area of buffer overruns. The new /GS flag, when code is recompiled with 71, will catch many buffer overruns. There are also Safe Exceptions which I won't go into here. Also, in the unmanaged area, there have been some outstanding improvements made in the optimizer, particuarly for floating point. The /G7 flag builds code optimized for the P4 and there are major gains. Also the new /arch:[SSE:SSE2] flags let you generate code for the Streaming SIMD and Streaming SIMD Extensions 2 instructions. Last, but certainly not least, there is now a Form Designer for Managed C++! You will be able to do everything you do with the C# or VB designer. You can visually design forms and work with Managed C++. This obviously opens powerful opportunities to do certain things like Direct/X without Interop penalities and much more. Its very exciting! 3:45:10 PM

The warmth and depth of this community never ceases to amaze me. I was just talking to Tomas down in Columbia, who has posted a very thoughtful reply to my essay), and Ingo at .NET One, (at the same time!) who are both encouraging me to go out on the circuit and speak on COM Interop. Since they asked, and Radio seems to have lost the pointers, here is the Is COM Interop Fundamentlaly Flawed (Parts 1 and 2) story which Radio has once again lost the links to, from last August. There is actually more detail in my slides on my web site. I will be updating this after talking to Don about it and with new insights but for now, Paresh, who worked heavily with me on this area during that period, has an excellent article on what we found in Connection Points in particular and how to write wrappers to fix some of the mess. 10:59:26 AM