>Since i really want to use Linux, i want some filesystem server for at
>least reading some kind of Linux fs (other than MessyLoss of course -
>i hate being limited to 8+3 chars in filenames). Do some other people
>are working on it now? (it seems that some guy started working on ext2
>but never got it done) I looked at several Linux fs sources (in Linux
>v1.1.60): ext2, xia, minix:
>
> 1. ext2 is (probably) the most used and the most efficient Linux
> filesystem. It is the most complex of the 3 fs. It is probably the
> most efficient, perhaps the less buggy (since it has been tested at
> thousands of Linux sites). It has cylinder groups, some kind of ACLs
> (partly implemented), some kind of interesting file flags (compressed,
> immutable) most of which are not implemented today (in Extfs 0.5a).
> Perhaps a possible advantage of ext2fs is that it has already been
> ported (according to source comments) to mach and/or hurd, masix
> which are supposed to be microkernels.
It's been ported to the Hurd (which runs on the Mach "microkernel"),
and is currently being ported to Lites (which also runs on Mach).
Both ports are by the same guy - ldd@step.polymtl.ca (Louis-D. Dubeau).
You might want to contact him - he probably has some good hints and
possibly some ext2fs code that's much more portable than the original
Linux code, which you could get going fairly easily on VSTa.
>I think i don't want to reuse code from Linux fs. I will only use the
>data definitions (ie the include files describing the disk
>layout). Linux fs code is full of linux specific kernel routines.
Reinventing the wheel will almost certainly take more time, especially
considering that ext2fs has already been ported to non-Linux environments.
>At
>the moment, i only want to read Linux fs. I don't want to spend to
>much time (ie only a dozen of evenings).
You can simply not implement the glue code to call the ext2fs write
routines, if you're in a hurry. It'll still almost certainly be
way faster than rolling your own from scratch.
Another approach you could take, if you're really not interested in
writing to filesystems even in the longer term, is to use the
FS interpretation code in the Mach server bootstrap program. It's
fairly well encapsulated and should be easy to rip out, and already
supports ext2fs, Minix, and BFFS, so you'd get "three for the price
of one". :-) But it's definitiely never going to support writing.
If you try this approach, you should coordinate with whomever is
planning to work on the new boot loader (Dave Hudson and/or Yossi
Gottlieb, both of whom are on this list), since it may use it too.
You might be able to use the same source code, or a common library.
Bryan
Received on Tue Nov 8 07:38:14 1994
This archive was generated by hypermail 2.1.8 : Thu Sep 22 2005 - 15:12:10 PDT