home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.software-eng
- Path: sparky!uunet!world!bzs
- From: bzs@ussr.std.com (Barry Shein)
- Subject: Re: Extent-based Filesystems (was: Large Application data sets )
- In-Reply-To: moss@cs.umass.edu's message of 23 Jul 92 13:27:18 GMT
- Message-ID: <BZS.92Jul26035303@ussr.std.com>
- Followup-To: comp.software-eng
- Sender: usenet@world.std.com (Mr USENET himself)
- Nntp-Posting-Host: ussr.std.com
- Organization: The World
- References: <1992Jul13.090423.20408@metapro.DIALix.oz.au <33120@cbmvax.commodore.com>
- <1992Jul21.113652.4898@metapro.DIALix.oz.au>
- <1992Jul22.194214.20451@sequent.com>
- <MOSS.92Jul23092718@ibis.cs.umass.edu>
- Date: Sun, 26 Jul 1992 08:53:03 GMT
- Lines: 56
-
-
- From: moss@cs.umass.edu (Eliot Moss)
- >It is certainly *possible* to related extents to structure. For example, some
- >IBM disks (in the earlier days anyway, not so sure now) were set up more like
- >rotating tape drives in the sense that the user could determine the lengths of
- >the blocks and where the physical inter-record gaps fell, and blocks did not
- >have to be all the same size. These were allocated in cyclinders and tracks,
- >of course, and a collection of standard formats were offered by the OS, but
- >one could come up with new ones I believe. But this no longer seems the way to
- >go on most systems; pre-formatting to a certain block size (and accepting a
- >little waste) is just easier, more portable, etc.
-
- If anyone wonders what is meant by "easier", in the old IBM OS/370
- days if you could not remember the formatting of a file and you were
- not a system wizard you basically could not open the file and would
- get errors and forced exits if you specified the wrong details in a
- job.
-
- I remember people deleting files and re-creating them from tape or
- cards or whatever because no one could find the job cards which
- described how the files were formatted on disk and the OS's reaction
- to a "guess" was basically to abort the job (unless the guess was
- correct, of course.)
-
- All files were opened in the JCL cards which preceded the job (again,
- by most mortals, from Assembler you could do anything if you knew how,
- but even that was difficult as the system call macros resisted certain
- types of hackery), IBM OS/370 wanted all files pre-allocated before
- the job ran because it used deadlock prevention (i.e. the job was
- pre-empted until all files etc were available) of a rather primitive
- sort.
-
- Although my memory is rusty a JCL card to open a file looked roughly
- like:
-
- //SYSIN dd dsn=foo.bar,lrecl=80,blksiz=800,space=(80,80,RLSE),
- disp=(new,keep,delete)
-
- which reads, roughly, open a file called foo.bar which later will be
- referred to by the logical name SYSIN (sort of like a Unix fd but
- symbolic, you could have numeric LUN's also), its logical record size
- is 80 blocked 10 per block (80*10=>800), and allocate space for (here
- I'm very rusty, help), 80 blocks, 80 more blocks if I go over that,
- and RELeaSE any I don't use at the end of the job, disposition is a
- new file, keep the file if I succeed (exit with R15==0), otherwise
- delete the file on any failure.
-
- There were, of course, about 200 other things you could say on that
- card (literally.)
-
-
- --
- -Barry Shein
-
- Software Tool & Die | bzs@world.std.com | uunet!world!bzs
- Purveyors to the Trade | Voice: 617-739-0202 | Login: 617-739-WRLD
-