Thank you very much François, I'll take a look at that and get back with you in the next week or so. For what it's worth I'll be on the #haiku-dev channel with the nick bjl-.
Cheers, Brian On Mar 3, 2010, at 3:31 PM, François Revol wrote:
Hi Brian,My name is Brian Landsberger, and I have recently become interested in the Haiku Operating System. As a Software Engineer in Test at Isilon Systems I recently built a general-purpose NFSv3 library in C+ + that used to create a userland NFS client, a torture test suite, and a PCAP data analysis application. While recently test-driving the operating system I found some bugs in the NFS implementation, and so I asked around on the Freenode irc channel about whether or not there was a need for NFS developers in the Haiku project. Although the bug I found was recently patched, it seemed as though the project was looking for people to contribute to a v3/v4 implementation. Please let me know how best to proceed.Sorry I didn't answer you directly. Indeed, the NFS implementation currently in svn is only for version 2 of the protocol, doesn't support caching, and probably still has some bugs. Which is why I was thinking about going for doing one for v4 from- scratch. IIRC, v4 removes a lot of old stuff from the protocol, and adds interesting things like a mount-less discover protocol, an uri scheme and xattr support (though the namespace isn't standardized). But if you have a working v3 implementation feel free to try and port it to Haiku. You'll want to have a look at existing filesystems, including the custom network fs Ingo wrote. You should be able to reuse your C++ code, as the kernel has C++ support (though no RTTI or exceptions, at least not by default IIRC).Using the userlandfs server you should be able to easily test the code.François.