Re: [seul-edu] Re: Best wm for tiny slow computers?

["Ralph M. Deal" <deal@kzoo.edu>]

Michael Shigorin wrote:
> 
> On Tue, Feb 19, 2002 at 08:26:28PM -0500, Ralph M. Deal wrote:
> > I just started distributing a 6.2 configuration that takes
> > about 360M but without the abiword editor I had hoped for.
> Look in /usr/share/{locale,timezone} -- some extra 20+ Mb could
> be reclaimed by deleting extra data.

I'll check that - I'm not at the site now.

> > tasks such as shutdown.  Any suggestions for an editor with
> > table and list capability smaller than abiword?
> If .doc compatibility isn't an issue, I'd recommend LyX.  It's
> more on the right way, as a side effect.  If you have network,
> you can typeset them on nearby teTeX; if you don't, then this is
> not the way probably.  And more on typesetting -- consider lout
> package if you'll get into it -- it's compact and should do the
> job (of typesetting).

I am assuming that most of these computers will go into homes
with very little computer experience, so typesetting is far from
their concerns.

> > I've had to copy disk images with dd by starting up and
> > shutting down LINUX between transfers since I must have cold
> what do you mean by "cold"?  Perfect copies are made with three
> disks in a system -- one to boot, one as fully-umounted original,
> and one more as a target.  Of course, there are variations as on
> booting from tomsrtbt, network, second root on original,
> temporary root on target...

I've been using dd ON the original/boot to copy the orginal and it
works!
That of course could be the problem;  if dd has to be cached on the hard
drive due to a small RAM, then the heads reading the original would have 
to constantly going back and forth.  Anyone know how dd works?
If dd is RAM resident, more RAM might help?  Are the data transfers
performed
directly or are they buffered in RAM?

> ...but if hassle is unbearable, "init 1; umount -a; sync; sync; dd"
> should do the job (although you get root partition copy
> non-clean in terms of fsck).

Think I'll stick to the dd transfers since it preserves the partition
structure well.

> > hard disk transfers.  Any better way (other than getting an
> > extended ribbon cable to serve several hd's at the same time -
> > how many are possible bythe way?)?  Would more RAM make that
> Depends on your layout -- doubt it has sense to use more than one
> active HD per IDE channel (so you end up with say booting from
> hda and then dd if=/dev/hdb of=/dev/hdc bs=1M), and that 75MHz
> PCs usually don't have more than two PCI IDE channels anyway.

I had been making the hd image copying with dd without any bs
specification
and it worked fine.  Would it be faster if I specified bs=1M?

> > transfer faster?  It takes about 16 minutes for the actual
> > image copying on the 16M, 75MHz machine.
> hdparm -d1c1u1m16a16 /dev/hdX (do read the manpage and experiment
> with one target drive first!) should set DMA on disks; you may be
> able to estimate transfer rates by running hdparm -t /dev/hdX
> (and compare them afterwards).  Check -a/-m parameters, as say WD
> drives are known for being broken in many places.

Very helpful.  I have read about hdparm in the Config-HOWTO but have not
yet tried it.  These probably are WD drives.

Many thanks for the detailed response!    Best,  Ralph   deal@kzoo.edu