On Thursday, August 26, 2004 11:55:24 -0500 Chris Green
wrote:

> Just a caveat from a (non-lawyer) lurker: the clean room approach may
> work if there are no patents involved (which protect algorithms, unlike
> copyright law). However, a previous poster to openafs-devel mentioned
> that the license covering the AFS code specifically mentions patents:


Of course, it helps to actually _read_ the license. Or, for that matter,
the post you referred to.

The IBM Public License contains provisions which require that contributors
grant patent licenses along with their contributions -- you can't
contribute code to OpenAFS and then later sue OpenAFS users over a patent
you hold on something the code you contributed does.


It is worth pointing out to people who haven't been following this thread
very closely, that the "clean room" approach to reimplementing something
doesn't really apply here. We're not talking about a piece of code that
implements something; we're talking about an _interface_.

What Volker's code does is implement the same RPC interface as the AFS
ptserver, which is responsible for name-to-id mappings and group membership
in AFS. His code is not a work-alike; it is a protocol translator that
turns pts lookup requests into queries to winbind. None of this was
borrowed from AFS.

The files he wants to include from OpenAFS constitute an interface
description -- they are used to generate the client and server RPC stubs
and associated headers, and to describe the interface that both sides must
speak.