Talk:Virtual file system

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Sun Microsystems[edit]

Any references to Sun creating the VFS? I seem to recall it first appeared in some form in Research Unix, but I can't find the reference for that either. Lost Goblin 05:50, 17 December 2005 (UTC)[reply]

Seventh Edition didn't have it; Eighth Edition might have had it, but it wasn't made widely available, and it came out in 1985, about the same time SunOS 2.0, the first SunOS with a VFS, came out. Guy Harris 09:17, 24 December 2005 (UTC)[reply]
I have heard from someone that was at the Labs at the time that PJW had a VFS layer in '84. VFS was certainly part of 8th Edition. But the exact date 8th Ed was released outside the Labs scapes me at the moment.Lost Goblin 19:21, 24 December 2005 (UTC)[reply]
Right, Sun invented the VFS name. See the paper given in the article by Steve Kleiman (later CTO of NetApp, alas, not the Steven Kleiman with article here). He gives a list of the team. (COI: Guy Harris and I joined the group later) The inside joke was the Virtual Node (vnode) was a pun on Vinod Khosla, a co-founder of Sun. The AT&T Unix variants called it a "File System Switch" but cannot remember if it was developed independently about the same time. W Nowicki (talk) 19:30, 9 June 2013 (UTC)[reply]

Implementation of single-file virtual filesystems[edit]

If we talk of files, which "internal structure [access] ... is often limited to programs specifically written", rather than filesystems written to support standard files API, then there should be much more examples.

  • Microsoft COM compound file, which examples are MS Office files, MS Management Console files, etc.
  • To some extent - MIME-HTML files and DBX files (e-mail storage by Outlook Express), where files can be accessed independently, but usually are not.
  • OASIS OpenDocument (technically it is ZIP with bunch of files, and Office suite works with those files)
  • Quake WAD files (at least in Alice McGee it is also a ZIP file) and other games' packages,
  • Sun Java .JAR files, that are some variation of ZIP too.

All of those formats feature having a set of data chunks with specified name, size and type. Essentially that is exactly what files are - named chunk of data. And if we remove requirement to make it accessible to any program via standard file I/O API and only request that programs, using them, do read and write data chunk by name, then all those examples above are valid and much better than the exotic WinUAE and VmWare. —Preceding unsigned comment added by 212.44.136.228 (talk) 08:16, 9 November 2007 (UTC)[reply]

To me, the whole section on single-file virtual filesystems should be deleted. Though these are "filesystems" and could be described "virtual" they don't mean the same things to software developers as "Virtual File System", which has a different, specific meaning, as described in the first paragraph. The technicalities of these systems are covered in detail under Loop device, and right at the end of that article there are a few details of implementations which are probably in enough detail. The above comments are not virtual file systems even by this nonspecific definition, as they are not designed to provide random access nor access for arbitrary software. They are just archives or structured databases. DanPope (talk) 12:31, 11 August 2008 (UTC)[reply]

Linux VFS limits file name length to 255 bytes?[edit]

Is this true that Linux VFS limits file name length to 255 bytes as stated in ReiserFS article? — Preceding unsigned comment added by 84.97.14.22 (talk) 12:39, 12 July 2012 (UTC)[reply]