Changes between Version 32 and Version 33 of SoC2013

Timestamp:

Apr 22, 2013, 3:46:08 PM (10 years ago)

Author:

Niall Douglas

Comment:

Added proposal for extending Boost.ASIO with async file i/o

SoC2013

@@ -182 +182 @@
 This project will be mentored by Christopher Kormanyos
 
+=== Boost.ASIO ===
+
+Boost.ASIO is probably the leading asynchronous i/o implementation library in C++, followed in popularity only by libuv which is pure C. Parts of Boost.ASIO are expected to join the C++ standard library in TR2 as the official C++ networking implementation (see N3360 [http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3360.pdf]). Yet something very obvious is missing: Boost.ASIO is missing asynchronous file i/o support for POSIX, yet there is a partial implementation for Microsoft Windows ([http://www.boost.org/doc/libs/1_53_0/doc/html/boost_asio/overview/windows/random_access_handle.html]).
+
+A very nice and useful Boost GSoC project would be to fully complete asynchronous file i/o support for Boost.ASIO and to submit it as a complete new Boost library (Boost.AFSIO?) which ties together Boost.ASIO, Boost.Thread, Boost.Filesystem etc. And in fact I have implemented most of such a library (see https://github.com/ned14/TripleGit/blob/master/triplegit/include/async_file_io.hpp), but the following items remain outstanding before it could be accepted by Boost:
+
+1. POSIX async i/o support is limited to a threadpooled emulation. Proper support for Linux's kaio and POSIX's aio_* interfaces would be great. Note that these APIs are extremely tricky to get working right across all POSIX platforms, especially OS X.
+
+2. The file i/o APIs supplied are the bare minimum. A deeper, more explicit integration with Boost.Filesystem which is currently only supported using hand written template shims would be needed.
+
+3. Documentation is rudimentary, even by Boost standards.
+
+4. Unit testing is not to Boost standards.
+
+5. The existing implementation is exclusively C++11 and does not use Boost's macro trickery for supporting older compilers. Some refactoring would be needed here to pass Boost peer review.
+
+6. The library needs to pass Boost peer review, with all the bureaucracy that involves. Do NOT underestimate the difficulty of this part - in fact, this part alone is probably 50% or more of a GSoC project. Yet it is probably the most valuable part of a Boost GSoC internship because if successful, your resume can state that ''you're the (co)author of a Boost library''. That's something only a few hundred people in the world can claim, and it will stand with you as a bright point for the rest of your career.
+
+This project would be (co)mentored by Niall Douglas ([http://www.nedprod.com/]).
+
+
 === Benchmark Library ===