Backup Success

The new WD "My Book" 2TB drive is installed and working. My life is backed up. My toilet isn't. Life is good!

Prior to the demise of the pocket-sized WD Passport that I used for backup until it quit last week (see post before last, below) after six or seven years of very satisfactory performance under not infrequently adverse conditions, I used GNOME's "Déjà Dup" backup utility provided with Ubuntu 12.04 LTS. Déjà Dup had several IMHO significant shortcomings, and I was looking to replace it: Déjà Dup produced specially formatted files that gave no access to individual backed‑up files except through... you guessed it... Déjà Dup. There was considerable debate in reviews of Déjà Dup, and even on the comment thread in the Ubuntu free s/w archive, about whether restoring individual files even worked right: some people claimed it restored an entire directory, possibly with disastrous collateral damage to files other than the one intended for restoration. Most annoying of all, there was no facility of any sort for browsing the contents of a Déjà Dup backup... the backup was just a humongous gigantic file. Not good!

Installing the WD "My Book" involved learning more than I ever intended to know about how Linux tracks filesystems. Why? Because WD provides tools for purchasers using a PC or a Mac. There's no problem seeing the drive or its contents in recent versions of Ubuntu Linux... the familiar Windows NTFS format works like a champ... but if you want the out-of-the-box main partition to auto-mount, you have to give some minor cryptic instructions in some files Linux knows to look for. Mother Web will tell you how to tweak those files, but I regret to generalize by saying that not all Linux adepts are the best writers of English in the computer biz, and comprehending exactly what to do takes a fair amount of browsing and trying. In any case, the one-and-only partition on this drive (as shipped) auto-mounts on bootup just fine now, and can be used just like any other. The thing is pretty fast, too.

So I moved on to the backup conundrum. The problem has two extremes: one is faced by sysadmin's with dozens or hundreds of machines networked in an assortment of ways and probably mixed on the network with PCs and/or Macs. There are expensive commercial solutions for that extreme. On the other extreme is the home user with one or a couple of PCs or Macs to back up. Apparently, WD supplies decent tools for that purpose which you can install from the drive. Unfortunately, there doesn't seem to be a WD backup utility for Linux.

After a bit of browsing and reading, I decided what I needed was not so much a formal backup tool as a snapshotting-synchronizing tool, something that copies all files in specified folders of your internal HD to the backup device the first time, then refreshes/synchronizes that backup each time you run it again, copying only the files required to accomplish that. That gives you the effect of a series of full snapshot backups, without the overwhelming disk space requirements that such a backup would impose. And folders and files look just like the originals... you can browse them with the most ordinary of apps, or, of course, navigate them with the snapshotter... it is its own browser.

In short, my choice was "Back in Time" ... a tool apparently originally developed for the Mac, then redeveloped for Linux by using several common system utilities wrapped in a convenient graphical interface. Serious Linux virtuosos might view it with contempt for all I know; people who can tap complicated scripts exactly suited to their own needs probably don't need Back in Time. But I am no script virtuoso, and on my worst days, I'm kinda lazy... so Back in Time will do the job for me. Tonight I tried it out for the first actual backup, backing up everything under /home/myusername. Even with two versions of all my photos and a bunch of other junk, it took well under 30 minutes for the initial (full) backup. Subsequent ones should be much quicker.

(Sorry; no pic of the drive. It's butt-ugly.)

The Right Tool For The Right Job... On The Wrong Workbench

This week, my ancient and venerable WD external HD, a cute lil' thing intended for my pocket when traveling with my laptop back in my working days, gave up the ghost. Though I am no longer a road warrior and don't even own a working laptop, I've been using the pocket HD for backup, so something had to be done about its demise.

Stella headed for Best Buy today to pick out a TV (one of those also almost quit this week; it will show some channels but not others... its been a bad week for electronics in Our House), so I rode with her, thinking to buy a similar pocket-sized HD. Given that I work (if you can call it "work")  exclusively at home now, I used that fact to liberate me from tiny and likely fragile devices, getting a much more ordinary external HD, again a WD because I've had good luck with them. I thought $99 for a 2TB drive wasn't bad. Hey, 2TB holds a lotta snapshots!

Instructions and setup s/w that comes with such things is inevitably of only two flavors: Windows and Mac. Use them with a Linux box? Suuuure; just be sure you know all the right config files to tweak manually. Linux these days handles NTFS filesystems with no sweat... again, if you know the right tweaks.

But I didn't. What should have been a 15-minute install took more like 3 hours because I had to learn how Linux thinks about filesystems to get all the permissions set right manually first, then change several boot files that do the same thing during a system start. I got it mostly right the second time I tried.

But as the venerable Troxel once said, the first 90% of the job takes 90% of the time, and the last 10% of the job takes the other 90% of the time, and so it did. (Troxel is an old colleague of mine who looks almost exactly like Charlie Pierce, and has politics almost exactly the inverse of Charlie Pierce. 'Nuff said.)

I rendered my system unbootable only twice, and only one of those times... the second... did I not understand the improvised fix with which I made it work. Everything is hunky-dory now. All I have to do is mess with all my backup scripts to reflect the change in drive name and folders...

Bang!

A transformer blew out with a loud report, leaving my block with a brown-out. At first I didn't realize there was a problem; I just reset a breaker and restarted my computer. But repeated attempts to connect to the 'net failed with modem problems. Eventually I noticed the lights were flickering, so I promptly shut down all computers... I once lost an expensive copier due to a brown-out.

About an hour later, the Kilowatt is Reddy again, and I'm back on the net. Some problems never really go away...

Open Software — Followup

Following up on the post below, I offer continuing adventures in open software...

I finally got the underpinnings of my project working, the whole stack of apps and components all connected together and (minimally) running. I don't feel like quite such a dolt tonight, though it may be a while before I feel comfortable developing apps for UNIX-Linux systems. One has to start somewhere.

In this case, my frustration was caused by problems in not just one but two of the four layers, the Tcl/Tk library from ActiveState and the PAGE visual form design tool. Unsurprisingly there were no actual errors in ActiveState Tcl/Tk 8.6, only a misunderstanding on my part as to how it was to be hooked up to the PAGE tool. And PAGE worked fine once I ran a configuration script that determined for PAGE just where it should look for its underpinnings. To complicate matters, versions of Python built in the past few years all have an earlier version of Tcl/Tk built in to the runtime/development environment, so that PAGE was attempting to talk to Tcl/Tk 8.5.11 rather than 8.6 . Both errors on my part were educational; I learned still more about some very basic *nix operations in chasing down the problems. I'm afraid I have a lot more very basic things to learn.

I also learned the challenge of installing and configuring software packages that do not reside in well-defined online repositories, which is how a lot of Linux software is distributed. The tools on Ubuntu (some of them borrowed from other distributions of Linux) do not make it easy to install a single package from a vendor who simply provides a single compressed archive (think: zip file).

One frustration to a newbie like myself is that in the *nix world, hardly anyone writing documentation... man pages, forum posts, introductory articles, whatever... feels a need to explain one single fact more than the minimal amount necessary for an experienced person (user or programmer) to use the software described. One painfully common fault is describing the changes one must make to a configuration file without ever mentioning where to find the file. Fortunately I found a good search tool; I'd have been unbearably taxed by doing all the searching in something as slow at searching as, say, Nautilus. This is a marked difference from the Windows world, where the doc writers' assumption is typically that the consumer of documentation is an utter idiot.

Onward into the fray...

Open Software: What You Get For Free And What You Don't

The First Computer Bug

Back in 1947, a year before I was born, when early computer wizard Rear Admiral Grace Hopper (USN) found a moth fouling a relay in the Harvard Mark II electromechanical computer, she removed the moth and taped it in the project log book, where I presume it is still preserved for posterity. In other words, it's not my fault: computer bugs have been around quite literally all my life. I'll be polite and not speculate on what Adm. Hopper said when the bug got into the relay; I assume that since she was Navy, she knew how to curse.

Geany

Tcl/Tk

Fast-forward (now there's an increasingly antiquated term!) to 2013, where I am in the middle of my n'th attempt to become even minimally competent at programming for a *nix (i.e., UNIX or Linux) system, n being a regrettably large integer. With great deliberation over each choice, I have chosen to learn the programming language Python (yes, it's named after Monty Python), the lightweight integrated development environment Geany (perhaps the only piece of developer software with which I've felt instantly comfortable; on the n-1'th attempt I used Eclipse on an old, slow machine and came to regret it), the graphical interface library Tcl/Tk (also known for historical reasons as Tkinter) and the visual form design tool PAGE. I don't know how common these things are out in the real world of *nix because I never had a contract in that world; I just have to start somewhere.

Things went swimmingly up to a point. I managed to construct manually a few simple programs using Tcl/Tk, tapping the code by hand, copy-pasting bits and pieces from examples provided by other people. That all worked pretty well.

Then, after the aforementioned deliberation, I set out to install PAGE. The author of PAGE claims it rests on version 8.5.4 (or newer) of Tcl/Tk. The version installed on this box by the original Ubuntu installation is 8.5.11, so I should have been OK.

I installed PAGE with no significant error messages, I attempted to run it from a command prompt. It quit with a complaint (presumably from the Python runtime) that the program (i.e., PAGE) was attempting to use calls to Tcl/Tk version 8.6, the latest version... NOT 8.5.4, as advertised in the documentation. Thus entered the dark side of free software.

ActiveState, one of the major commercial software and support companies in the Tcl/Tk world, offers a prebuilt freebie version 8.6 package of Tcl/Tk, with no support. I can see the advantages of their doing so: programmers that grow up using the freebie version could very likely influence the subsequent purchase of commercial licenses by their employers or clients. Freebies are pretty much traditional in the Linux world, and the market of people who use them would be a good market to tap into.

I downloaded the Tcl/Tk 8.6 freebie package and attempted to install it... three times, using three different install tools. I got hundreds of error messages, which I could not redirect to a file (due to my ignorance, I'm sure). I ran a few of the handmade programs using 8.5.11 to make sure I hadn't destroyed that; apparently it's OK. But after an entire evening of my life spent on this project, I am back to square one. Even for a guy who used to do this stuff for a living, that's pretty frustrating.

And like Adm. "Amazing" Grace, I know how to curse!

Back Online In My Own Chair

I now have a working Ubuntu Linux 12.04.2 LTS machine, a probably 5-year-old PowerSpec retired when I retired, functional and effectively free, but at a cost: the install DVD (downloaded .ISO) insists that it could not install Ubuntu Linux so it could cohabit with Windows XP. It may be something was wrong with Windows (it was virus-ridden; that's why I retired that already somewhat old computer in the first place), but I had to sacrifice that copy of all my photos. I have several backups if I can only find them. That's one of many things I need to look for. Oh well, at least this box is noticeably faster... Blogger's editor no longer drags down; both processor and memory are adequate to the task.

I did my exercises today. I can really feel the effects, positive and otherwise. Daily exercise is a foreign notion to many programmers, especially old guys like me. But if I want to stay alive and walk on my own two feet (one natural, one purchased), I have to do it.

Time to look for a few more missing items for the computer... sigh!

Steampunk Economics!

MONIAC, Bill Phillips
(via Timothy Taylor)

Via Paul Krugman, we have Timothy Taylor talking about an hydraulic analog computer, MONIAC, built in or before 1949 by Bill Phillips (of "Phillips curve" renown), intended to model an economy in the presence of user-specified changes to fiscal and monetary variables. Taylor in turn points us to information about an even earlier economic hydraulic analog computer built by Irving Fisher in 1891. (No, I did not transpose a digit: I mean 1891.)

Technophiles and lovers of steampunk will appreciate these devices for their own sake; I, at least, am reminded once again, not for the first time, that there was effective practical mechanical computing of very complex functions of many variables long before there were digital computers of any sort. Economists and similar researchers may learn from the existence of these machines that, as Krugman emphasizes, it is important always to have a model, in the broadest definition of the word, of any system about which one proposes to argue: the lack of a model inevitably leads to certain types of errors that the very process of building a model (physical or mathematical) forces one to deal with a priori. I don't know economics very well, but I can attest to the truth of Prof. Krugman's statement in other contexts: modeling is essential to larger problem-solving.