A package was submitted to savannah.nongnu.org This mail was sent to [EMAIL PROTECTED], [EMAIL PROTECTED]
J. Buchanan <[EMAIL PROTECTED]> described the package as follows: License: gpl Other License: Package: Secure, Scalable Unicode Filesystem (UniFS) System name: unifs Type: non-GNU Description: There is no source code yet for this project, why will become clear in the description that follows. UniFS (may be renamed GnuFS after consulting with rms, in progress) is a new free filesystem that scales to petabyte-sized storage devices, supports Unicode internally, is based on extents rather than clumsy indirect blocks, supports strong cryptography, compression, secure deletion of files, file slacks and directories, supports the creation of encrypted swap space for the host Operating System and may end up as a journaling filesystem. A requirement of the filesystem is that it is self-healing and crash-proof. The filesystem is organised into layers, with a *BSD and Linux compatible VFS and Gnu Mach/HURD compatible server interface (which hooks into the VFS layer, followed by the UniFS-specific layer, then a device-independent disk-driver layer, then a layer of glue code that interfaces directly with device drivers on the host Operating System.) For compression, the UniFS will rely on GNU gzip, for strong cryptography, the GnuPG-derived libgcrypt (found on the GNU alpha server), for character set translation GNU iconv, and for the secure deletion, GNU shred (part of GNU fileutils/coreutils). Designing a filesystem that is for the most part Operating System and machine independent is a mammoth task. In particular, ensuring that the filesystem reaches its goal of scalability to petabyte sized secondary storage may be impossible in the near future (but it will be tested on terabyte sized storage.) The architecture and specification of the filesystem is near complete, but coding will not commence until issues with the strong cryptography have been sorted out, for example storage of symmetric keys encrypted with public keys on disk extents or elsewhere on another filesystem, and how application programmes take advantage of this; and there is the issue of bypassing kernel page/buffer caches or implementing the filesystem's own buffer/page cache. Here is a brief overview of the filesystem architecture: [VFS/HURD server layer] Provides mount/unmount, read/write, fsync, open/close, ioctl, etc system call stubs. [UniFS layer] Provides the filesystem functionality: on-disk structures shielded by the above platform-independent layers (VFS and HURD server interfaces), cryptography and compression, secure deletion, information needed by stat(). [Machine independent block I/O Layer] Provides platform-independent interfacing to platform-dependent device drivers. This layer must bypass the Operating System's virtual memory and cache mechanisms. [Device driver layer] Provides "glue code" to interface with the actual disk drivers, and possibly the filesystem's own buffer/page cache. In porting the filesystem, only this layer and the top interface layer need be changed. Initially documentation on the UniFS will be released (within the week). Articles and references will also be supplied and requests for comments/suggestions will be sought. No developers will be accepted by the project until it begins the coding phase. Other Software Required: None that aren't already part of GNU, though it may use Diet libc instead of Gnu libc. Diet libc is free software too. Other Comments: I have sent rms details of the filesystem and await his comments, and I'll keep the Savannah administrators updated on any progress. _______________________________________________ Message sent via/by Savannah http://savannah.gnu.org/ _______________________________________________ Savannah-hackers mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/savannah-hackers
