Re: "bugfree :)" ramdisk-1.3.4.tar.gz now available

Chad Page (cpage@best.com)
Tue, 27 Jun 1995 18:01:37 -0700 (PDT)


On Tue, 27 Jun 1995, Drew Eckhardt wrote:

> IMHO, a ramdisk block device is the wrong approach to getting a
> memory based filesystem
>
Yes - a true memory-based fs would be a better solution. I've
been thinking about doing one, but haven't gotten around to it yet...

However, there are some benefits to a ramdisk block device that
is dynamically allocated (Drew's points are better though :) :

1. It allows work to be done on rootdisks without actually
having to allocate a disk partition, and very quickly (a loopback driver,
such as the one done by Ted can do this as well)... it also allows easy
preparation of compressed rootdisks, the other use of my ramdisk code.

2. It is not as complex to write (?) as a tmpfs would be. VFS
is very well done, but it still would take a while to do.

3. A ramdisk image can be moved around and copied very easily,
and placed onto floppy or hard drives... file fs's can be tarred, but you
can do it a little more directly on a ramdisk.

> - It allways uses a fixed ammount of memory, irregardless of
> how full it is.
>
Actually, a filesystem created in buffer cache will take up only
the disk space actually used - unused parts will not be allocated until
they are accessed...

> - It uses a "normal" filesystem on it, with the overhead associated
> with that.
>
I think even a memory-fs would have it's own traffic, but I see
your point...

> - You need to mkfs any ramdisk you want to use. PITA.

That dosen't bother me too much - it is a very fast process when
done on a ramdisk.

> instead, we want a generic memory based filesystem; which would
> eventually be pageable. We should do something along the lines of
> Sun's tmpfs...

Agreed... but in my opinion both should have uses.

Something else I would like as well would be to have a VCS
(Virtual codec system). This would allow gzip compressions, as well as
many other kinds, to be handled in kernel space. Filesystems could than
use those codecs to compress and uncompress data in a consistant manner.
I'd like to see file compression (as done on NT 3.51) on Linux. Volume
compression is much less stable IMO.

- Chad