[ros-dev] Re: FAT patents again... Global and High Performance File Systems

[D. Hazelton]

On Sunday 12 March 2006 06:01, art yerkes wrote:
> On Sun, 12 Mar 2006 10:31:39 +0100
>
> nitro2k01 <nitro2k01 at gmail.com> wrote:
> > I'm not talking about invalidating the patent as of today, to favor eg
> > ROS, but I'm rather asking how they got could file the patent? Didn't
> > IBM react? And moreover, was it legal for them at that point of time
> > to file a patent regarding something that already existed. (Namely a
> > file system capable of storing a long and a short file name for the
> > same file)
> >
> >                                                /nitro2k01
> >
> > On 3/12/06, Oliver Schneider <Borbarad at gmxpro.net> wrote:
> > > > Reminds me... didn't OS/2 have the capability to store both short and
> > > > long file names on a FAT16 partiton? How does that go with MS'
> > > > patents? Where MS really allowed to file a patent on something that
> > > > already existed?
> > >
> > > I think the actual question is how you want to invalidate the patent.
> > > As far as I remember it costs even money if you want to attack a patent
> > > that you think is prior-art. How does that go with the work of
> > > volunteers...? I am involved in other FOSS projects and I would not
> > > want to sue someone else on behalf of a project or to be sued instead
> > > of the project. At last it is a volunteer's work ...
> > >
> > > Just my 2 cents,
> > >
> > > Oliver
>
> It seems that the consensus was that a vfat mode that stores only
> long filenames would sidestep the patent.

AFAICT, the patent is valid because of the method it uses. OS/2 used a 
different, though somewhat compatable method.  Now the fact is that a VFAT 
mode that just stores the long filename does sidestep the issue...

However, the FAT system, while simple and easy to implement, does present a 
lot of security problems. A better solution would be to build a new 
filesystem, perhaps using the FAT system as a base and have it incorporate 
everything people think they need.

This could mean file versions, it could mean journaling, definately means 
ACL's and advanced permissions and file owner information is another 
must-have. 2K and above also offer per-file/per-directory/per-drive 
compression and encryption - I have not seen how this works with 2K on an FAT 
system, but with NTFS it's quite seemless, even if it does cause some hiccups 
in the load time.

I'd be more than willing to help design the system, though my skill at such 
low-level coding for operating systems other than Linux is limited. For now, 
though, I'd suggest just keeping the FAT system in place - the MS patent only 
covers how the LFN's are stored, not the basic architecture of the filesystem 
itself.

DRH