The licensing concerns should not be problem because there is a precedence to this and Oracle is already aware of it. The port from LLNL is under US government grant and has been validated by their lawyers. Obviously at first sight it seems like there is problem with the CDDL/GPL issue but as is clearly demonstrated by the port by LLNL and validated by their lawyers porting it as a kernel module is not a problem.
see
http://www.osnews.com/story/23416/Native_ZFS_Port_for_Linux
Have you personally or someone from your company validated that with Oracle themselves? The source you indicated is nothing more than rumour, and not evidence that Oracle is fine with the situation. Their lawyers may be waiting for solid evidence of licence infringement before proceeding.
Does LLNL research grants and liability-cover extend to other 3rd parties such as yourself? (I wouldn't believe so, particularly due to International borders being involved and I don't know how the US works, but here liability cover only covers the immediate grant holder, not 3rd parties involved).
Reading the LLNL site for ZFS on Linux. The ZFS components build as a kernel driver/module, but due to licence restrictions,
the resulting Linux Kernel and ZFS kernel module can NOT be redistributed. The end user is required to build their own kernel and matching ZFS kernel module. (Source:
http://github.com/behlendorf/zfs/wiki/FAQ ). All LLNL provide is the source code and instructions.
How does what you are offering get around the issue of redistribution of binary code? (which I assume you are doing). Or do you also require the end user to self build a kernel and kernel module which any sane business model would not support?
Also you still haven't mentioned what version of ZFS you are implementing? And user-space integration tools? GNOME integration? etc.
Also how does your work relate with what LLNL are doing? It is shared research/development? Is someone copying technology/research/development from another source?
At the moment, all your comments just open up more questions... which for me just builds doubt in what you are offering.
PS. If someone want's ZFS why just run Solaris 10 (which is fully supported by Oracle), OpenSolaris, Solaris Express 11 or FreeBSD where they are implemented without licencing issues and have a support base to rely on? What you are proposing appears to be a high risk adventure with what you little information have presented.
If Linux was a requirement (and I don't know why - since user space applications matter more these days than just the kernel), then wouldn't BTRFS be the storage technology to look out for? If it hardware drivers - without being too pompous - select hardware off the Solaris HCL and be done with it. And if you truely know kernels, then what does the Linux kernel offer that the Solaris kernel does not? (I'm having trouble understanding the business case for this adventure).
There is a group that uses the OpenSolaris nv134b kernel with a complete GNU environment - no SYSV stuff anywhere - The Nexenta Project. (
http://www.nexenta.org/ ) if you really need a complete GNU environment? And it's supported! No hidden catches. And you get ZFS.
PPS. I currently run Solaris 10 as my main desktop OS (yes, I'm a masochist), but I have StarOffice 9, OpenOffice 3.2, Firefox, Thunderbird, Acrobat Reader, Flash 10.1, Java, nVidia drivers for graphics and all the applications I could hope for (including games). Could you please tell me what Linux has, that I don't have or have access to?