Article I wrote for my local user group
I wrote the following for SLUG : Sydney Linux User Group
mailing list. I thought this list might find it of interest.
I hope that it doesn't disrupt what list members are used to.
Regards
Jamie
------------------------------
Recap
=====
Last installment, yours truly was waxing lyrical about a kde +
debian potato install.
I'm still pleased, but not all is sweetness and light.
First, let me backtrack a bit, and tell you what my aim is.
Goals
=====
I'm interested in the state of usability of distributions, given
some starting points:
* I wanted to see how far I could get on bottom range cheap
hardware. My goal is to see if a whole lab at a school or
similar could run Linux.
* I wanted to see if the proposed Linux disto could be used
for simple word processing, web browsing, email. I have as
subjects two daughters (10 and 12) with school assignments,
hotmail accounts, neo-pet habits and possible tetris withdrawal
symptoms.
To that end, what I was proposing to do was try this on hardware
which would have been normal three to four years ago. In my case,
this is a Pentium 133, 1 Gig drive. I started with 16 Meg of
memory, but bumped this to 64 Meg. S3 video card. I started
with a 14 inch monitor, but I think this is too restrictive, I
bumped this up to 17 inches.
I already have a 486-33 with a 200 Meg hard-drive, 16 Meg ram and
14 inch monitor. This is quite popular, as it is reliable. However
it is very slow and has limited functionality. It runs a funny
RedHat 6.2, with bits of X and Netscape 4.
I intend to repeat the experiment using the equivalent of an X terminal.
The kind of hardware I would use for this may require a faster
(100Mbit) Ethernet connection, and I would like to have a completely
fan-less system (i.e. noiseless).
Implicit in this setup is a server system and a firewall. The server
runs samba, is attached to a printer, and runs the dhcpd daemon
(allocates IP addresses to workstations). The server is also
available for NFS. When I run the X terminal, the server will run
applications for the terminals. The server will eventually be a reasonably
modern machine.
In the back of my mind, I envisage this setup as being able to be
supported remotely, through an ssh tunnel. In my mind's eye, I imagine
a support person handling a number of sites remotely, with an
occasional visit.
Why kde and debian?
===================
Well I can't be comprehensive and cover every disto. I'm interested
in hearing other peoples opinions here, but my goals lead me to believe
that distos that tend to be no-brainers to install want very modern
hardware : all PCI, gobs of disk.
Also, I get the impression that easy to install tends to mean
hard to tailor, but am willing to listen to other's experiences.
That dropped out Mandrake, Progeny. I was already familiar with
RedHat, but wanted to try something different.
I was impressed by Kde on another system I had seen, and wanted to
give it a go. I had heard excellent reports of Konquerer. Kde
seemed to tie together well.
Impressions of potato
=====================
The installation of this was as I expected. The use of dhcp
certainly made life easy. In addition, I fully expect that
this installation would be easy to replicate on a number
of similar machines.
Potato + kde
============
This also installed very nicely, and was certainly very responsive
with the given hardware.
Konquerer lives up to its reputation. This is more than a useful
browser, and deserves respect (as Ali G would have it).
The torture test.
=================
Ok, the assignment is due tomorrow, the 12 year old has taken
control. My name will be mud (well it is anyway for countless promises
of "I just have to install it" : the tinkerers curse) if the
words are not on printed paper by ten o'clock. It is now seven.
The twelve year old has never seen Kde, Kword and I say "go to
it". I know I have to get printing working, I go to another
machine and ssh back to her system to try to configure it.
Kword certainly looks the goods. It actually has more capability
than is required here, with frames, style-sheets and tables.
Typical users don't bother with style-sheets.
Kword was certainly easy to use. By eight thirty she had produced
her report, complete with a table.
Printing was a bit of a problem. I haven't quite gotten to the bottom
of things (in panic mode, rational resolution gives way to expediency)
but the traditional lpr/lpd mechanism is not a friend of the dhcp
dynamic IP environment.
On the client, ip address to hostname resolution seems to be
required, either through /etc/hosts or dns, neither of which
seems to be straight forward with dhcp. On the server side, /etc/hosts.lpd
seems to want the ip address / name of any machine that is able to
print. Documentation is sparse, I don't know if the server lpd
can take ip+mask addresses in the hosts.lpd or if every ip address
in the domain has to be listed individually.
I'm not sure what the answer is here, I briefly explored rplr, and
probably will get to LPRng or cups, but am unfamiliar as yet.
My panic mode solution was to print to a file and use smbclient
(the samba server was already configured), as in
smbclient //server/lp -c "print file.ps"
Tales of woe
============
Up until this point, things were under control. Just when you
think it's safe ....
Kword is very easy to use and quite powerful. I mentioned before
that it has style-sheets and frames, features which push it into
the more serious category, and were a bonus for my target audience.
I was only looking for some quite basic functions. Tables would
probably have been the minimum feature to acceptance.
This version of Kword, and I really must check which version this is,
has some horrible formatting bugs. Changing fonts in tables changes
the screen, not the font that is printed!
In addition, the K ghostscript viewer was broken, we couldn't check
what would have been printed without printing.
Ok this is Linux, let's try abiword or ted.
Both these are simple word processors that take rtf.
Abiword wouldn't read the rtf generated by Kword. However ted would.
Ted, though, has a terrible tendency to crash at the drop of a hat.
In the end, we got the assignment printed using ted.
Resolution
==========
Despite the panics, I'm reasonably happy with the way things are going.
Kword had problems, sure, but these could easily be just some simple
bugs, perhaps an early version, or perhaps a simple workaround.
The ghostscript viewer might simply be missing some fonts. And
the printer support is probably only an apt-get away.
I wanted to say more about my Debian experiences, but I want to
hold off for now. I feel I could be generalising from a limited
exposure. I want to explore this system more.
One thing that has crystalised in my mind is the poor integration
of standard Unixy ways and dynamic ip addressing. I highlighted
printing here, but there are other examples.
Luke was posting recently about hooking up optushome. This is a
system that wants you to take your domain name from their dhcp
server. You pick up your mail from mail.their.funny.domain.name,
you use their proxy at proxy.their.funny.domain.name.
This probably works OK when a single host is directly connected,
but with a shared connection and firewall, I don't like it.
Conclusion
==========
I happened to hear on the radio yesterday, Gerry Harvey of Harvey
Norman bemoaning the slump in PC sales, currently 13% on last year.
He expected it to last beyond Christmas. He also saw large changes
coming in the future.
Harvey is no technologist, but he is a savvy box mover. The signs
of change are present.
The cost of traditional boxed software is now a very large component
of consumer PC's. These PC's are extremely frustrating for normal
users when they go wrong, and expensive to maintain.
There are many situations where PC's are not appropriate computers.
In addition, I sense a feeling that the computing needs of users is
plateauing, not needing expensive processors, nor software with
an extremely long list of features. In my example case, a huge
difference was made by adding more memory and a larger screen, not
by upgrading the motherboard and processor. In addition, the word
processor features were more than adequate with Kword.
This plateauing was denied by Gerry Harvey, ("he would say that,
wouldn't he").
However, a simpler, cheaper, quieter device may be just what
the public wants. It may prove more attractive than old loyalties.
Who will provide such a system in the future?
Reply to: