PDC 2009 Day 2: Microsoft Perspectives on the Future of Programming
This PDC session was a little different from regular PDC sessions as this was not composed of a speaker and a presentation, but is composed of a panel of experts instead. This panels was composed by: Butler Lampson, Erik Meijer, Don Box, Jeffrey Snover, Herb Sutter and Burton Smith.
I wasn’t present during the entire session, but here are my highlighted quotes from the part I was present in (unfortunately the quotes are brought without the name of the speaker who is quoted and I apologize for that):
- What is your view on transactions?
- “Transactions are like fairy dust. Programmers write their code as usual and than just sprinkle transactions around and everything works!”
- “Software tends to work when it works and fail when it fails. Transactions are a wonderful way to handle situations where failure is possible”
- “Don’t wait to use transactions. Just do it.”
- Can type systems help us write better code?
- “A type system is a Neanderthal program verifier.”
- “Type and memory safety is the way of the future, and any interesting language is going to have these systems.”
- Should pointers across machines (distributed pointers) be explicit?
- “Boundary transitions should be clear and known to the developer”.
- “The boundary transition should be clear since the connection to the pointer destination can fail, as opposed to the path from cache to RAM which rarely fails and thus transition boundaries there are not known to the user”.
- “URLs are actually remote pointers”.
- “I lung for the day that URLs are treated as first class language constructs”.
- Should we leverage the GPU cores for computational purposes?
- “GPUs are going to take form of more general purpose processors in the future, but we’ll probably going to call them GPU’s for a long while”.
- “We’re going to see much variety in processor architectures specialized for certain purposes and GPUs are going to be one of these types. The question is how do we utilize all of these CPUs in a uniform way”.
- “Developers don’t really want multi-core processors. They want faster single cores”.
- Should we have garbage collection in C++?
- “Having garbage collection is really really important… What was the question?” and in response: “His memory was garbage collected while he was speaking”.