home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ogicse!usenet.coe.montana.edu!uakari.primate.wisc.edu!ames!news.hawaii.edu!spectra!oles
- From: oles@spectra (Shawn Oles)
- Newsgroups: comp.os.linux
- Subject: Re: Stacker like file system
- Message-ID: <1992Aug30.024915.24276@news.Hawaii.Edu>
- Date: 30 Aug 92 02:49:15 GMT
- Article-I.D.: news.1992Aug30.024915.24276
- References: <BtpF3M.2xn.2@cs.cmu.edu> <1992Aug28.223823.9755@uoft02.utoledo.edu>
- Sender: root@news.Hawaii.Edu (News Service)
- Organization: University of Hawaii, Dept. of Electrical Engineering
- Lines: 54
- Nntp-Posting-Host: spectra.eng.hawaii.edu
-
- In article <1992Aug28.223823.9755@uoft02.utoledo.edu> billp@jupiter.cse.utoledo.edu (Bill Parquet) writes:
- >mfp+@CS.CMU.EDU (Michael Polis) writes:
- >: |> Having a /usr/compressed/pics directory with on-the-fly compression would be
- >: |> a dream come true.
- >: How about storing them as GIF's, which are inherently compressed using
- >: the LZW algorithm (the one used by compress).
- >
- >There's a few disadvantages to using GIF format. Most namely is the
- >restriction to 256 colors.
- >
- >What kind of compression technique do those
- >programs [stacker/doubledisk/etc] use? (Hopefully this isn't the start of
-
- I am not sure about the other companies, but stacker is based on Stac Electronics
- QIC-122 compression standard which is briefly discussed in "The Data
- Compression Book" (Nelson, 91 pg 228). Basically it uses a variant of the
- Ziv-Lempel Dictionary discussed in their 1978 paper and exploited by Welch
- in the algorithm LZW. There are a wealth of books on data compression. Read
- the FAQ in comp.compression for more detail.
-
- >a completely unrelated discussion) ... The only thing that I think gets
- >really ugly about those programs if you have a file that contians a block
- >of zeros in the middle, then you have a program that fseeks to that block
- >and writes a block of random data over it. The compression program has
- >to deal with expansion in the physical file size.
-
- easy. If blocks are compressed at the block level, as we have been discussing,
- the new block can be re-written in the middle of the file independently of ALL
- other blocks in the file; thus just the modified block, with its new size, is
- written. This not as easy as a normal modify but is doable.
-
- This does lead to a few problems, most notably, if the file system is full and
- the modified block size increases, it is possible that the file system
- will have to generate an "no space on device" type error message, which
- might break a few programs.
-
- One place where this worries me slightly, is what happens if
- a) if the file system is full
- b) and the user deletes a file, which modifies the block containing the
- files directory entry
- c) and the modified directory block expands (which is highly unlikely, but
- possible)
-
- if this occurs, then the user will not be able to remove the directory
- entry, and might lead to a corrupted file system.
-
- Anyway, the easy solution is just to panic("your screwed\n"); and recomend
- a reformat. :) This is a highly unlikely senario, but still bothers me.
-
-
- -shawn
- --
-
- -shawn
-