In praise of floppies

Behold the floppy, a miracle of post-modern technology! Does your computer still have a floppy drive? Count yourself among the lucky!

This is a floppy.

Why? Thousands of reasons! Do you have an ultralight lifestyle? Are you a maximalist? Are your computing needs sufficiently sparse to fit your entire home directory into less than a megabyte and a half?

If so, then you too can use floppies to back up your personal files! And if you are daring, you could even run your entire system with a floppy as the home directory. Yank that floppy out, and bingo! your precious data are safe and secure!

Even better, with two machines with floppy drives, you can synchronize systems in a fraction of the time it takes to string cables, set up network interfaces, configure wireless keys, install a networking protocol, configure a server and client, get them handshaking, manage groups and permissions, and then synchronize. Floppies save you time!

Floppies are cheap! Chances are, someone will give you a floppy — how many times has someone given you a USB drive that didn’t include some sort of blatant advertisement, or preinstalled crapware? No crapware on floppies!

And floppies are versatile too! You’re not limited to using an out-of-date, obsolete, archaic filesystem with floppies. You can use ext2, ext3, ext4 and even the newest, coolest filesystems on the block! Don’t want journaling? With floppies, you can skip journaling altogether!

Floppies are disposable, and sometimes even recyclable! Tired of your floppy? Make a clone of your floppy, and move it to a new one in seconds! You can decorate a floppy with colored markers! You can slather it with stickers — but not the puffy ones!

Floppies are available in thousands of colors, dozens of styles and themes. If you’re lucky, you may even find — gasp! — the holy grail of floppies: the transparent floppy!

Floppies are universal! You don’t need a proprietary interface or a conversion cable or a rubberized palm guard to use a floppy. Do you have a floppy drive? Congratulations: You can use floppies!

Are you a Linux guru? Do you want to be a Linux guru? You’ll dazzle them at your Linux guru job interview by mentioning that you always install grub to a floppy, so your computer is unbootable without it. It’s like a primitive boot lock!

And don’t forget, you can still install one of the greatest operating systems in the world … with only floppies! That’s right, you can get your system online and surfing the Web with only a few floppies and an hour or two of time. It’s magic!

Yes, there’s still plenty of use for floppies. Don’t let angst-ridden teenage geek wannabes and turtlenecked pseudo-minimalist artsy-snob-types tell you otherwise: If you’re not using floppies, you’re just not cool!

Paid for by the Committee to Re-Elect Floppies.

An Android in the family

I’ve mentioned in the past when family members get Linux-driven computers, so I suppose I can mention this as well.

Astute Reader Number One sent this picture the other day of a Samsung Fascinate (I think that’s the proper link), which is new to the household. I know almost nothing about these fancy phones, but I do understand this to be an Android-based gadget.

Maybe in about 15 years, when they’re out of date and completely unsupported, I’ll get one and string it into the rest of my archaic home network. It can’t be worse than this.

Be prepared to do it yourself

I have a bad habit of giving advice on running open source projects, which is something I have only the slightest measure of experience with.

But I’m going to do it again, so brace yourself: Be prepared to do everything yourself.

And by that I mean money, time, effort, skill, artistic talent and even advertising, in the sense of increasing public awareness of your project.

I read yet another high-minded site this morning, suggesting quite a few lofty changes in a particular package, changing this and that and pointing out flaws here and there. (No, no link this time.)

But every suggestion was punctuated with an “I can’t do this alone, though,” or “I don’t have time to take care of that myself,” or even “I need someone to do this for me.”

Sorry, but I have almost no hope of seeing those changes ever take place. You’ve shot yourself in the foot already, and you’re not even past the imagination stage.

Maybe I’m a pessimist, but suggesting changes or additions and then saying you can’t do it, that someone else has to get their hands dirty. … Well, it’s not just silly, it’s almost rude.

Be prepared to do everything yourself. You have to be sufficiently motivated and talented and equipped to go from zero to 100, from version 0.0-alpha to 1.0-final, under your own power.

If you get lucky, someone who is like-minded might join your troupe and complement your skills. And that makes things infinitely easier. But prepare for that as the exception, not the rule.

Because if you’re not ready to tackle every aspect of your project, no matter how daunting, then you’ll end up two years down the road with a stale blog, a couple of mockups and nothing much to show for it.

Don’t wait for someone to do things for you. Do it yourself, or don’t do it at all.

Surprised, and not surprised: Feather Linux at 150Mhz, 32Mb

Some things surprise even me. And yet, at the same time, this isn’t a surprise.

 

I was racking my brain a few weeks ago, trying to remember a distro I had tinkered with probably four years ago, when I was looking for a suitable low-end distro for yet another low-end wonder.

I don’t think I made any mention of it here, because if I search this site for “Feather Linux,” not much comes up. And to be honest, outside of this site, I don’t see much mention of it either.

Fact is, given the opportunity to mention a lightweight distro, most people jump straight for the jugular and shout “DSL” or “Slitaz.” Not that there’s anything wrong with either of those.

But something low-end-but-not-wicked-sparse like Feather Linux probably never got the benefit of the doubt. I am sure there are some 200Mhz machines out there that would benefit greatly from Feather, but are force-fed DSL apace because it’s “light.”

If I understand the home page correctly, Feather started as a Knoppix remaster, with a little of this and a little of that, and matured all the way to version 0.7.5 — with kernel 2.4.27 and Firefox 1.04 — before it became a ghost distro.

And that’s a shame. Honestly, this runs magnificently at 150Mhz and is only using a fraction of its 32Mb to show the desktop. I can start most programs without heavy swapping or losing system response, provided they’re not overweight to start with.

Boot times are not stellar, but are relatively quick. In a world where Gnome Ubuntu setups on reasonably fast but slightly out-of-date machines can take a full three minutes, two minutes for a Pentium to load Feather Linux is an achievement.

And the variety of software at your disposal is amazing. Everything from Abiword to Siag to Sylpheed to Elmo to XMMS to e3 … and much more. I cower in awe.

All that in 128Mb. Just the “Net” submenu is a veritable who’s-who of Linux network software. But I won’t pretend Feather is god’s gift to the ancient computer; it has a few shortcomings that are noteworthy.

First of all, kernel 2.4.27 did things a little differently than what I’m used to now. I have this machine on a very reliable wireless card, but can’t get it on my in-house network to save my life.

Similarly, sound is a mystery, although I daresay I could solve that in a matter of a few minutes, if I really wanted to. I feel I have enough information and experience now to pin down the audio with a few experimental commands.

But I could say that about any distro, from aLinux to Fedora to Salix to ZOMG Linux. No matter what you use, you should be ready to tweak a few things, to make it your own.

But Feather Linux is something I feel confident in recommending. Depending on your hardware, and depending on your ability level, and depending on your patience, this could be only a few steps away from regular use.

It’s out-of-date and esoteric in some ways, but that doesn’t mean you can’t put it to use on your closet-dweller. Try it out.

gcc the pig

My 133Mhz Pentium does three things. First, it is a round-the-clock torrent slave. I drop in torrents over NFS, it downloads them automatically. I copy the results back.

Second, it’s a local “mirror” for the Crux repositories, and in turn serves them to two other Pentium machines in the house. Rather than tie up the repos with each machine syncing individually, two sync off the third.

And last, it compiles software updates on its own, on a nightly basis. Those compiled packages in turn get fed to the other two Crux-driven Pentiums, which saves me time and effort.

Setting it all up is a breeze. One three-line cron script washes out the port tree, syncs with the official servers and starts the updates.

I had to adjust things slightly a day ago though, because one particular package was causing problems: gcc.

gcc the pig was taking well over 24 hours to compile, and bogging down the machine so badly I could barely sign in to kill it off. It had swallowed up the vast majority of the processor and memory, and most of the swap space too.

I should have realized that some of those packages, although there are very few, would take quite a long time to build if they were updated. So I am partly to blame.

In any case, a quick --ignore=gcc added to the prt-get sysup command solved the problem. Happy days are here again.

The also-rans

I’ve mentioned a few candidates for the 150Mhz, 32Mb machine, which I keep on hand to deliberately confound with software written 14 years after its creation.

A few things have proved usable, but only in the slimmest, narrowest definition. Slitaz is the big winner, of course, and the age-old TinyFlux got a small nod. And of course, an outdated Ubuntu console.

But not everything gets a smilie. Here are a few that I tried but didn’t get much out of. For reasons that will be explained. …

I don’t remember how I came to find TurboPup, but it was a bit of a challenge to install. Installing Puppy Linux, for some reason, occasionally confounds me.

After fighting with emulated systems and installing from live USB drives, I finally got one TurboPup system written to an image, transferred it across, and started it up.

And it claims it can’t find its critical pup_420.sfs file. Oh well, I tried. Maybe a straight installation of Puppy, instead of a backdated derivative would be a better idea.

As a side note, if I could suggest one thing to the people who make Puppy Linux, it would be to cut down on the wordy dialog boxes. Pretty please.

I don’t want to speak for everyone, but it’s a bit tedious to skim through two paragraphs of narrative to find that there’s really only one button that matters.

It’s not a crime to put one terse sentence beside an option, and use a help button to explain in detail.

Vector Linux, like I said before, wanted a seemingly immense 6Gb or so to install properly. It went against my grain but I went ahead and gave it that space, with the intention of moving the system across by way of a full-size external drive.

I was expecting a transfer time of over three hours when it moved across the Mebius’s USB1.1 port, which is pushing my limit. But it never came to that, because the installer kept freezing after the partitioning sequence. I took the hint.

I should mention that I did try Mandriva too, but not really with the purpose of pushing it onto a Pentium machine. Doesn’t matter, because it wouldn’t boot for me, either in an emulator, from USB or from CD. I must be doing something wrong. (And yes, I verified the checksums.)

Since Ubuntu 10.04 refused to boot, it shouldn’t come as any surprise that Linux Mint Fluxbox built from a recent Ubuntu also refused to boot.

LMF relayed the same error messages about “real mode pages” and “loading a kernel first,” in its failed efforts to satisfy. I hold no grudge here though.

Because honestly, this was a bit of a dream. My previous efforts with Linux Mint cast from Ubuntu have told me that I really should be seeking out lightweight distros, not just distros that use lightweight window managers.

What I could really hope for now, is for some enterprising person to take the Linux Mint Debian system and trim it down to bare essentials. Now that might have hope at 150Mhz.

That’s all for now. I have a couple more candidates on the list, and the image transfers only take up about an hour at a time. So my search certainly isn’t exhausted yet. And neither is the hardware.

No two are alike

Never feel frustrated or intimidated if something I suggest doesn’t work. For every one successful note I have here, I usually have three or four giant question marks.

Sometimes those question marks are interrelated, which makes them all the more frustrating.

For example, I have a terrible time trying to remember how to power off the backlight — not just blank the screen, but to turn off the backlight — on my console-only framebuffer systems.

As a result, and because I get tired of trying various options only to rebuild a kernel and waiting for it to (maybe) work, I get lazy and just start closing the lid on a laptop. The BIOS usually turns off the backlight for me.

But at the same time, most graphical X systems can turn off the backlight. Somehow. Magically. So as a troubleshooting measure, I decided to try building X on the 133Mhz Pentium file server and torrent slave.

It should work just as well as the Mebius, I figured, since they’re both Trident video cards and use identical software to display a graphical screen. Neither can do vesa, both are 800×600. Et cetera, et cetera. …

Except that the slower Pentium gets a whitewash effect with the trident driver. It almost looks like the screen is being over-charged or something, with a slow melting white smear that shows a faint shadow of a distended pointer.

Same card. Same generation. Same kernel version. Same Xorg version. Same package, same labeling, same compilation flags.

But the fbdev and vesa drivers, of course, don’t work, even if I recompile the kernel to support a framebuffer — which I originally didn’t.

So why does the Mebius get a proper 800×600 graphical desktop (with Musca and rxvt-unicode) under Crux, but the 133Mhz Pentium with comparable hardware doesn’t?

Beats me. I just work here. If I had all the little answers I could spend a lot more time chasing the bigger ones. One day I’ll get the backlight issue solved, and it won’t entangle me in other, stranger problems.

Of course, another one will come along. …

Tale of two Ubuntus

Two of the systems I planned on trying out with the Mebius were Ubuntu versions, one four years older than the other. Results were what I anticipated, although there was a small surprise attached.

The console version of Ubuntu 10.04, as I expected, wouldn’t even start. My experiences earlier this year foretold that a machine with a meager 32Mb of memory wouldn’t get past the grub menu.

And that was the case this time too. Ten-point-oh-four left me with a few nifty error messages (“error: cannot allocate real mode pages,” and “error: you need to load the kernel first” were common) and a feeling of inferiority.

Dropping back four years, just out of curiosity, worked fine. The installed system booted to a console prompt and needed something less than 11Mb to run. The wireless network was up with a little prodding, and all was well with the world.

Adding a graphical environment was a fruitless exercise though: The Trident driver that far back gave nothing but scrambled eggs for video output. TinyFlux had its artifacts, but a minimal X in Ubuntu Dapper was pointless.

Oddly enough though, a full-blown Ubuntu 6.06.1 desktop could handle a graphical environment neatly, with no obvious defects or artifacts, showing the deep brown background from ages gone by.

Unfortunately that’s all it showed. Anybody can tell you that a full Gnome desktop on 32Mb is nothing but a pipe dream. The machine was swapping before it ever snapped out of the console.

Eventually it just spun to a crawl, freezing with the drive light fully on. As expected.

Not much was gained here in this little experiment, and I acknowledge that. It does however reinforce a few experiments from a long time ago, when I started out with minimal systems and built them up.

Next up, a few systems that might actually have a chance of working.

Back in time: TinyFlux at 150Mhz, 32Mb

TinyFlux is something I tried a long time ago, when it was still active, and tried again last week because I thought it might work all right at 150Mhz on 32Mb of memory.

And my hunch was more or less correct, even if there is such a small margin for error with this machine that the slightest bit of excess baggage tips it from functional to failure.

This particular version of TinyFlux was branded as specific to the i586 and used kernel 2.6.18.x, with a smattering of software that — while technically out of date now — is still very useful. Provided your hardware can keep up, of course.

I installed this the same way I worked with Slitaz, by running the ISO through Qemu and writing to a disk file, then using dd to write the disk image to the flash card over USB1.1. Not the quickest way, but it does work.

 

Startup went fine, but the network and video were botched — the latter I automatically blamed on the video card’s inability to work with the standard vesa driver that most distros use these days.

It was a different problem this time though: TinyFlux, in its installation procedure, had written out the /etc/X11/xorg.conf file to point at the cirrus driver, which it was correct in doing … under the emulator.

So switching to the actual hardware meant the wrong driver was in use. A quick hand-edit of the configuration file, substituting the “trident” driver for the “cirrus” put it back on its feet.

Networking was likewise easily solved, just by inserting the yenta-socket module. After that, the orinoco-cs module is automatically inserted, and TinyFlux’s network discovery and configuration put me online.

Like Slitaz though, TinyFlux is quite sluggish at this level. A clean and fresh system rides around 16Mb of 29Mb available — which is quite good for a fully graphical system — but almost instantly the machine begins to rely on swap space.

Opening the Fluxbox right-click menu causes swapping. Moving the arrow through the menu causes swapping. Clicking the selection causes swapping. And after that it only gets worse.

Opening Netscape Navigator to its home page takes more than 20 minutes, and that’s with absolutely nothing else running. It’s about 10 minutes before the toolbar appears, two more to see the page area painted, and much much more before the home page at wikidot.com appears.

And all that time, the drive light is unflinchingly lit.

Dillo and some other lighter alternatives are on here, and I can say that with only one of those running, you can do much more in less time. Honestly I didn’t expect Navigator to actually do anything at 150Mhz.

But to be honest, it’s more than just slow processor and limited memory. I’m getting a lot of graphical artifacts on the desktop. There are small, miniature corrupted windows tiled along the bottom.

A lot of the styles and wallpaper are probably too intense for the machine to keep up. The entire desktop seems a bit overwhelmed at times. Glossy buttons and rollover effects are redrawn visibly.

And like I said, responsiveness deteriorates quickly after the system starts up.

That being said, and as I mentioned two years ago, it’s definitely not ugly. It’s a good looking Fluxbox desktop with enough shortcuts and menu options to keep the most rabid Fluxbox fan sedated.

So yes, I like it, I just wish it could work a little more cleanly at this level. It’s a lot of fun to see a full-fledged distro come to life on a 12-year-old computer, but beyond the starting desktop, there’s not a lot I can get done.

And yes I say that knowing that this machine is still further below the practical threshold for this particular distro. Give me a 10-year-old K6-2 with 128Mb of memory, and this would be one of the first things I would suggest for it.

As it stands though, I shall continue searching. My hunch wasn’t wrong, it was just a little off the mark.

Three steps to instant cool

I seem to be getting a lot more traffic than usual these days, which is either a good thing or a bad thing. If you’re new here, welcome, and I’ll summarize much of the content for you. Just as a courtesy.

Here are three things you can do to take an old, unwanted computer and turn it into the coolest, sharpest toy in your neighborhood. Follow through with these steps, and every geek — even the turtlenecked wannabe artist types — will bow down to you.

1. Take it apart and repaint it. I know, it’s easier said than done, particularly when we’re talking about laptops. But you’ll learn a lot about the guts, you’ll get a chance to clean everything out, and when it’s done, it’s absolutely spectacular. Don’t skip this step.
2. Swap the hard drive for a CF card. Whatever you heard about CF cards or even SD cards being untrustworthy or having a short life span is utter and complete rubbish. Forget you heard it. I’ve been using one for about three months now, and spent almost three weeks raging against another, and they both still work fine. Get one. Now.
3. Install fbterm. This is also easier said than done, but in doing so you’ll carve your system demands down so drastically you can easily get by with a 10-, 12- or 14-year-old computer, and it will do everything you ask of it without even blinking (except maybe the cursor ). And best of all, it’s insanely awesome and cool beyond measure. You’ll be so glad you did this.

That’s all. If you need details on how to get started, I’ve kept all my notes and tips scattered around the site. Take a week or two to get used to your machine, then get to work revamping it. I promise: The results will turn people green with envy. …