Darwin's Theories - OpenBSD category

KDE4 - First Impressions

Sat, 02 Feb 2008 23:28:13 GMT

As most readers know I use the UNIX operating system, particularly OpenBSD, on all my computers. On desktops I flit about between several "desktop environments" or user interface systems. OpenBSD's default is a simple window manager called fvwm, but it also provides an extreme minimalist window manager called cwm. When I want a full-function GUI, I use KDE, one of the two mainstream UNIX desktops (the other is GNOME). I find KDE more intuitive than GNOME, and I also like the fact that its default configuration only has one icon/menu panel (GNOME has two, which is a nightmare on the 1280x800 laptop that is my main desktop), and it's at the bottom, with a start menu at the left (like that other OS from Redmond, which I have to use at some client sites). Linux founder Linus Torvalds agrees with my sentiment; his "Just Use KDE" comment a while back attracted a lot of attention.

I've used KDE 3 for several years on OpenBSD, so it's natural to expect a new major release. And indeed, KDE 4.0 was released in January 2008. And (thanks to work by Marc Espie) KDE 4 is now working on OpenBSD (although, because it's so new and because it can't easily co-exist with KDE 3, KDE 4 will probably not be included in the upcoming OpenBSD 4.3 release).

KDE4 is nothing if not innovative. The new desktop shell, Plasma, tries to move away from the old desktop-as-directory paradigm using more dynamic features, which you can read about here. It supports widgets, similar to those on Mac OS X but which can be written in any of half a dozen modern languages, not just JavaScript. The graphical effects are also quite impressive. Plus, it's faster at some operations; opening a file chooser on a large directory used to be painful, but is now quite fast.

So why do I say "almost ready for prime time"? First, I'm seeing some instability - Plasma disappears every so often and has to be restarted, for example, or kwin crashes and gets restarted - but these might be because the KDE developers are only coding for Linux, not for portability (though parts of KDE4 - not including Plasma - can also run on MS-Windows and Mac OS). However, more significant is that the system just doesn't feel "finished". Many of the optimizations and features (especially configuration) that we took for granted in 3.5 are not in 4.0 yet. For example:

The "pager" feature that lets you have multiple virtual desktops. It seems to work almost exactly as in 3.5, with the annoying exception that you cannot get the page miniatures to display their names, only their numbers. The UI for "Configure Desktops" lets you assign names, but the UI for "Pager Settings" has only a checkbox for "Display the Desktop Number", nothing for "Display Desktop Name".
Start Buttons. I like single buttons to start common applications. KDE3 had a very nice Konsole (terminal program) control applet for the task bar, for example. The "widgets" that have taken the place of Panel Applets currently consist of a much smaller set (8 or 10 instead of 30 or so), and don't appear to be installable on the panel. As it turns out, for programs, you can find the program in the menu, then right-click, and select Add to Panel. And for Widgets the same as I found out by searching forum sites: you can drag Widgets directly from the Add Widgets dialog to the Panel (but not from the Desktop to the Panel), and not all Widgets yet work correctly in the Panel. For programs, you can also select Add To Favorites, which puts it in the first section of the K Menu.
For Programs or Widgets installed in the Panel, you have far less control over how the panel is laid out. This may be a conscious choice (see Decisions); if so it flies in the face of the desire for ultimate configurability.
Gratuitous changes to Konsole (the terminal program; I must admit that at heart I'm sometimes a command-line guy) mean that my "terminal striping" - commands that put the system I'm logged into, and the current directory - don't work. I presume it's just a different set of escape sequences, but why change it?
Clock - can't have a sub-list of selected timezones, have to pick from the entire list every time you move (painful for road warriors).
Kopete, the Instant Messaging client, doesn't have IRC support (even though it did in 3.5.8, and even though the Kopete developers hang around on an IRC channel named #kopete :-); the code has only had one commit in the last month.
Migration Wizard - there isn't one, though one has been discussed at the Ubuntu Forum.

The first release of anything as big as KDE is bound to have some issues, and it does. Others agree with me; see KDE developer Aaron Siego's blog, and also this poll (when I viewed it on 2008-02-03, results were 12% "Works Great", 48% "Works but buggy", and 40% "still too buggy"). I'm hoping that 4.0.1 will correct some of these, and that the rest will come without too much delay (4.1). I quite like KDE 4, have been using it for a few days, and look forward to being able to deploy it as my standard desktop when it's fully ready.

Thecus Back To Life

Mon, 26 Nov 2007 23:47:00 GMT

The Thecus N2100 whose untimely demise I blogged recently is now, I'm happy to report, back to life, thanks to some help from the folks at Thecus in Taiwan. I'm putting some finishing touches on the OpenBSD install notes that should be ready in a few days.

IPv6 or Bust, and, Farewell IPv6 Samurai

Wed, 31 Oct 2007 16:33:00 GMT

In case you've been off-net for the last ten years, you should know that the Internet is running out of IP addresses - the telephone numbers of the Internet - at an alarming rate, and will reach the failure point within three or four years. IP is the Internet Protocol, the software that figures out how to get your Web, Email, VOIP or other data from computer A to computer B through a changing mesh of other computers, routers, and links. The designers of the current version of IP - IPv4 - were good designers but never anticipated the commercial success of the internet technology, nor did they envision internet-connected televisions, heating/cooling controls, and home networks in every home. The result is the impending doom of "Sorry, you can't have Internet service, we're all out of addresses."

The stop-gap solution of Network Address Translation (NAT, which some Linux people rebrand as "IP Masquerading"), only goes so far. Many of the protocols involved do not work over NAT. This is not the answer.

The long-term solution is IPv6 (longer article on IPv6). Don't ask what happened to IPv5; for that matter, don't ask what happened to IPv3. But I digress. IPv6 allows 128 bits of addressing, compared to IPv4's 32 bits. IPv4 thus allows just over four billion connections, and the Internet has estimated three billion already, growing about 170 million per year. IPV6, however, allows trillions of trillions of addresses. Write the number one (1). Now write seventy seven zeroes after it. That's how many connections IPV6 allows, more or less. While there are already more people on the planet than IPv4 addresses, there will never, ever be 1.15 * 10^77 people on this planet. Should our insane population growth bring us to 10 billion people, the survivors will be able to allocate 1.15 * 10^67 IPv6 addresses each. If they still have electricity to run their computers, that is.

So, how do we get from IPv4 to IPv6? That has a lot to do with politics, economics, and technology. First, we know that the herd mentality leads humanity to keep running in the wrong direction even when the cliff is pointed out to us, trying to veer away only at the abyss. Just look at how we are continuing to mis-manage the environment, the oceans, the future of the planet. Second, end users don't know or care about the problem, so they will not push their ISPs to support it. And some ISPs will say it's not their problem until users start asking for it, though some of the more sensible ones already have it up and running.

Sadly, the reality is there may not be a push to adopt IPV6 until the year 2011, when people can't get service. Then, of course, it will be done hurriedly, and poorly, and with no thought for security. But there are some pushing for its adoption; in fact IPv6 has been available for almost a decade (as soon as the Internet started becoming popular, some of its stewards realized that the address pool was too small). My own experience is with OpenBSD, the leadingly secure UNIX O/S, which has supported IPV6 for years. A man called Itojun - officially Jun-ichiro Itoh Hagino - worked tirelessly within a project called KAME to prepare a V6 implementation that would work in the BSD family of operating systems. The Kame IPv6 is now in OpenBSD, NetBSD, FreeBSD, and even Mac OS. Similar code is in Linux. And there is support in modern Microsoft systems. If you want to geekishly try it out before your ISP has it available, there are several "tunnel brokers"; the one I'm using is SIXXS.

As recently as a few months ago, Itojun pointed out that the current IPv6 standards committee had re-introduced a stupidity from IPv4 known as "source routing". Imagine the long-distance abuses possible if you were able to dictate to the phone company that "I want this phone call from New York to Boston to go via London, Istanbul, Beijing, Hawaii, but bill me only for the New York to Boston trip", and the phone system obeyed you without asking for administrative confirmation. Source Routing basically allows you to do this with Internet data. Itojun constructed a (carefully controlled!) test that was able to bring part of the internet to a standstill for a few seconds by routing traffic around and around and around...

So why do I mention Itojun, when the IPv6 story is much bigger? I found out yesterday that he passed away in Japan at the age of 37. Others are better able to eulogize him ( bugtraq, undeadly, sixxs) than I, who only met Itojun once or twice at OpenBSD gatherings. But he will be missed, and an important voice in the move towards saving the Internet from itself has been forever silenced.

Flying Thecus has hard landing...

Fri, 15 Jun 2007 01:21:00 GMT

The Thecus N2100 sits idle in a corner, because the vendor changed the software in a way that breaks the ability to install most open-source OSes.

The OpenBSD install, like many others, depends on setting certain firmware commands to be saved in the flash memory, and executed at boot time. This worked nicely on the versions of Thecus that the OpenBSD support was developed for, but in current versions (1.93+, including mine) Thecus modified the firmware to still save these commands, but to ignore them at boot time and use their "standard" ones. They also removed the source code of previous versions from their web site, making it very difficult for even advanced developers to figure out how to revert this unfortunate change. They have also refused to provide information on how to fix it, saying I must contact my reseller (since I bought it used, there is no reseller). I have been told that other open source projects are having similar problems, so it's not just OpenBSD that's affected.

Until Thecus changes this code back, so it actually honors the fconfig boot script settings, I can not advise buying this unit. If you want a small standalone unit, consider Soekris instead.

Flying Thecus Eats Cereal, err, Gets Serial

Thu, 14 Jun 2007 02:35:00 GMT

The Thecus I wrote about previously is partly up and running. The OpenBSD install, unlike that for Debian, requires use of a serial console. The serial port was designed into these devices primarily for debugging, not what a consumer appliance needs. So nowadays the manufacturer saves a few pennies on each unit by not soldering the internal connector in place. But at least the holes are there. The older units have a 10-pin header (with one pin removed) on the back of the disk circuit board. The newer ones only have a set of 9 holes for you to solder in your own header. Since mine came sans header, I just installed one. I used a right-angle header since the ribbon cable I had was straight, so the angled header leaves more room for air flow at the back of the unit. Connect a ribbon cable, plug the other end into a computer's serial port, and boot; you should see some textual chatter. Right? Right in theory. But not necessarily in practice. In fact, it's about 50-50...

The ribbon cable that connects from the header on the disk board to the serial cable or the serial port on the computer is of the same type as used on older i386 PCs, but there are two different types of 10-pin IDC to DB-9M ribbon cable in use, which look identical (the differences are hidden inside the DB-9 connector). Trust the "pee cee" industry to devise two totally different and incompatible cables and not provide a standard marking for them. Details on these cables have been committed to the OpenBSD installation document for the Armish.

So now I can see bootup messages, and interact with "RedBoot", the firmware boot loader these machines use.

RedBoot(tm) bootstrap and debug environment [ROM]
Red Hat certified release, version 1.93 - built 17:25:00, Feb 6 2007

Platform: THECUS N2100 (IOP80219)
Copyright (C) 2000, 2001, 2002, Red Hat, Inc.

RAM: 0x00000000-0x08000000, 0x0004b890-0x07fd1000 available, total: 128 MB
FLASH: 0xf0000000 - 0xf1000000, 128 blocks of 0x00020000 bytes each.
== Executing boot script in 3.000 seconds - enter ^C to abort
^C
RedBoot>

The next step was to install a hard drive as per the vendor documentation and OpenBSD as per those notes. These both went smoothly. The last step will be to make it automatically boot up when powered on...