Re: [Fwd: squid fs question]

Heya,

sfs was a good idea, but development on such has been pretty much
non-existent, for a number of reasons. So it doesn't really exist in a
functional way - at least, not in any way that Squid can use it.

However, the answers to your questions are yes, we store inodes instead of
filenames - the exact details on what gets stored where inside squid were
changing as the interfaces between squid and it's filesystems changed; and sfs
manages blocks allocated to a file by keeping a bitmap of allocated blocks
combined with using the first half of the first block (and potentially other
mapping blocks if the file is really large) for traditional inode-ish information
(ie. block pointers).

KJL

>>> Henrik Nordstrom wrote
> > Subject: squid fs question
> > Date: Wed, 10 Oct 2001 16:00:32 +0200
> > From: "Uritsky, Sasha" <Suritsky@ndsisrael.com>
> > To: "'Henrik Nordstrom'" <hno@hem.passagen.se>
> >
> > Hi
> > Sorry for bothering you, I am looking for some information concerning squid
> > filesysem. I've found a short description of it in the web which says, in
> > particular, that sfs is 'referencing files by their inode rather then by
> > name'. I've also found the names of its developers, Stewart Forster & Kevin
> > Littlejohn, but I was unable to find neither their emails nor email or
> > mailing list of sfs project. So maybe you can help me with connecting to
> > those people or simply by answering the following questions:
> > 1. given the above description of sfs, does that render the internal use of
> > directory tree virtually unnecessary?
> > 2. how sfs manages blocks allocated to file?
> >
> > really thanks
> >
> > sasha
>

-- 
Internet techie                    Obsidian Consulting Group
Phone: +613 9653 9364                    Fax: +613 9354 2681
http://www.obsidian.com.au/           darius@obsidian.com.au