lkml.org 
[lkml]   [2000]   [Aug]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: NTFS-like streams?
Date
On Sat, 12 Aug 2000, Michal Jaegermann wrote:
>Proponent of Rube Goldberg-like contraptions for storing "extendent
>information" about files claim that an obvious answer of using
>directories for such purpose is "brittle". A long real-life experies
>from NeXTSTEP does not support that.
>
>On NeXT various "NeXT Apps" are simply directory tries and command line
>tools do not even try to hide that fact. If you are using graphical
>file tools then by default they will show up as a single entity but an
>extra command, not hidden very deeply :-), will allow you to open such
>unit as a directory and get to components separately. This is,
>actually, a _good thing_ as a user can, for example, adjust and
>customize various interface elements using Interface Builder. After
>many years when NeXTSTEP is around I do not recall serious complaints
>that this leads to a breakage and if somebody will acutally mess
>something up then things are easily fixable.
>
>OTOH various other complicated schemes for storing a meta-information
>invariably turn out to be a major PITA and _really brittle_ outside
>their very narrowly defined "native environments" and break regularly
>when somebody will do something "unexpected" (read - use some standard
>and universal tool).
>
>With all this evidence in mind I watch pretty amazed when this
>discussion periodically flares up and various kludges are proposed
>as if nobody tried something like that in the past.

I agree. Trying to do an FSCK on such a beast must be REALLY fun.
The old UNISYS had similar things to track file revisions. Half the
time the file got really screwed when there was a crash - the same
line of fortran would be repeated (once for every change made to it).
It also made the filesystem very slow.

The only support I can think of would be a flag on a "directory" to
indicate to an underlying file system that the file is complex. That
way when a unix tar, cpio, dump ... (they have to carry the bit) copies
files out, the mkdir would recieve the flag. The underlying FS could
impliment it as either a simple directory (plus flag), or a complex file
with native streams. Tar and cpio copying from a filesystem with the
native streams would see a directory with the bit.

No changes to open or anything but the directory functions to pass
the flag back and forth.
--
-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@cats-chateau.net

Any opinions expressed are solely my own.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:58    [W:0.075 / U:0.100 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site