Re: Sharing Models, was: Intestinal Fortitude

From: Alejandro Dubrovsky (s335984@student.uq.edu.au)
Date: Thu Mar 22 2001 - 09:00:31 MST


* Robert J. Bradbury <bradbury@aeiveos.com> [010323 01:40]:
>
> On Thu, 22 Mar 2001, Alejandro Dubrovsky wrote:
>
> > Just a small nitpick: raw access has been around for over a year. And now
> > there are multiple journaling file systems (3 in my count).
> >
>
> No quibble, except that now I have to evaluate 3(!) journaling file
> systems to determine which of them may be the correct one to adopt.

sorry, i missed one. so there are four that i know of:
ext3 - extension of ext2 and backwards compatible with it so probably the one
people will most likely adopt, with all the associated advantages and
disadvantages of ext2.
reiserfs - journaling filesystem specialising in high speed and efficient access
to lots of little files
jfs - ibm's journaling fs
xfs - sgi's journaling fs

>
> However, the lack of raw I/O, a concept over 26 years old, and
> one that I personally implemented the use of over 17 years ago
> in the dominant RDBMS in the world today, in an operating system
> that has has all the hype associated with it that LINUX has had
> points out the lack of knowledge of the people doing the 'hyping'.
>
> But I'm glad to hear that it is finally getting the features that
> robust systems require.

i don't think it was ever an issue of wether it could be done, but the fact that
linus didn't want it in the kernel (i'm ignorant of the reasons of why he gave
in, or the reasons for the original resistance (apart from the fact that it
didn't fit the nice "everything over the VFS" scheme)) but i think the patch
existed for a long time before it was taken up so you could say that linux has
had raw device support for at least a couple of years.

Also, any RDBM vendor that felt like having raw access could have distributed a
kernel module with their product giving them raw access to the devices they
required. Probably not a major task for an RDBM developer.

Alejandro



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:06:36 MST