Re: [boost] sorting library proposal (Was: Review Wizard Status Report for June 2009)o

Subject: Re: [boost] sorting library proposal (Was: Review Wizard Status Report for June 2009)o
From: Vladimir Prus (vladimir_at_[hidden])
Date: 2009-06-03 03:47:31

Edouard A. wrote:

> Are you guys saying that all algorithms in boost are currently proven?

I believe that Boost libraries that have non-trivial algorithms
use established ones. For example, consider Boost.Graph -- where correctness
of operation cannot be established by testing, and where complexity estimate
is paramount. 
 
> If I were to submit an I/O library (it's an example, I'm not working on I/O)
> using somehow a "new" algorithm to manage asynchronous requests, would you 
> ask me to use a more classical approach during the review?

To make the discussion more specific, can you give pointers to classic
algorithms for managing asynchronous requests and outline the differences
you propose?

If your example is made up, then I can come up with a made-up answer. Say, you
want to support priority for async requests, and you decide to use specific
flavour of heaps for that. It would be fine. However, if you also invent your own kind
of heap, never seen in literature, that would be somewhat suspect.

> What is the difference between a new algorithm and a new implementation of
> an old algorithm?

Is this a genuine question? 

- Volodya