[boost] yet another asio review

From: Jose (jmalv04_at_[hidden])
Date: 2005-12-30 13:57:21

I vote YES to accept asio into boost.

Below are some general comments, mostly on what I think should be added.
Many thanks to Chris for sharing his expertise and continuing to work on
this much needed library !

I have an FC4 x86_64 box and have spent several hours using the library and
writing basic examples. I have spent many more hours surveying networking
libraries and ensuring I wanted to use asio

- documentation and examples
It would be nice to add some client/server protocol guidelines that help in
designing protocols.  Something along the lines on what Chris wrote to
compare using the reactor pattern vs using the proactor pattern in asio but
geared to different protocols characteristics like fixed header vs variable
header, line-oriented, .. The documentation could also include some tips on
how to handle special cases raised in the list, like throttling ..

An HTTP GET client example would be nice. Not a complete example but maybe
an skeleton that can be compiled and shows how asio should be used. No doubt
this would be the most used example !

- design

This is the main reason I became interested in asio. The design is not only
sound but also reflects the practical and wide networking experience of the
author

- implementation

I think the library should use Boost.Threads maybe in addition to the
current approach. By using Boost.Threads with asio we'll ensure continued
feedback/improvements for two libraries that in many cases will be used
together

- performance

While it has been raised that a synch iostreams interface is nice for basic
usage, the major value of C++ asio is performance. I think
it would be useful to keep some focus on this, maybe by providing a basic
client framework to measure server performance.
A basic usage for this would be to compare/benchmark different asio server
implementations (latency and throughput).

Look forward to the new releases !!