Archive for January, 2004
Linkblog
I need a link blog. Some place to store all the random links that I would otherwise bookmark on one of my many browsers and machines. It’s generally going to be public, but there may be a category or two that I’d rather keep private.
The most important thing is that it has to be easy to post from multiple locations/browsers with as close to one click as possible. I think that this means a RESTian interface for a javascript bookmarklet. The blogger api would work on the mac, since there’s the RadioService application that will post selected text from any cocoa based application. But that’s not going to carry over to Mozilla or the windows box.
It has to be accessible from all of my machines and likely locations. This seems to indicate that it’s going to need to live somewhere other than the laptop. This rules out Radio Userland.
I’ve got to host it. I’m guessing that it will wind up on the linux box, but the iMac server is a possibility too. I can use either filesystem, embedded db, or postgresql storage — MySql is out for now since I don’t already host with it. On the other hand, it could be a good learning experience.
Minimalistic is good — I’m expecting posts to be Link + Description, and maybe a Title. I should be able to get a lot of that from portions of the document with a bookmarklet. Google needs to be able to index the public categories. Don’t need comments, blogrolls, trackback, heavy duty navigation, much tempting, or anything terribly complicated. I’d like to keep it valid xhtml so that I can play with some of the more interesting searching, but that might tend to indicate that I want to be serving at least the public facing side of it using flat files.
I’d like to be able to edit online to add categories, but the permalink needs to be category independent. If I have to fire up bbedit to edit on the local machine, that’s not too bad, so long as I don’t have to muck with permissions. This is the drawback of approach #1, Blosxom + the RadioService + the Blosxom Metaweblog plugin. Since the webserver is writing filesystem files, they’re not going to be of the right permissions for me to edit them.
Blosxom is pretty close to the right solution but it’s not quite there. Knowing me, I’ll wind up using tiny bits of a massive system to do this.
No commentsCar Salesman
Well, this explains the car salesman thing.
They aren’t supposed to shut up, and they do make things up. And it’s mostly the fault of the highers up.
Resurrecting a Cube, Take 2
Trying this again, with feeling this time.
- New PRAM Battery. No Change
- ‘reset-nvram’ in Open Firmware. No Change
- Hitting the CUDA button to reset the power manager. No change
- Disconnecting the internal drive. Still can’t boot 9.1 on the iPod. I can boot a firewire external 10.3 drive, which I didn’t try before.
- Trying a fresh install of 10.3 on a (firewire) partition that has never had OSX on it. No change. It seems to have trouble writing the startup disk setting though.
- Zapping PRAM a bunch of times. No change
- Tried removing one stick of memory. Left it with only 64 megs, wouldn’t boot at all.
- Tried removing the other stick, left it with 128 megs and it booted X but not 9.1.
- Tried blessing the 9.1 system folder. Still nothing.
I have uncovered that the powerlight issue is probably due to something in the manufacture of the computer, as the serial number indicates that there was a fix applied afterwards and models of this age need to be sent to an Apple center to be modified.
I’m pretty sure that the rest of it is due to either bad firmware or a hardware problem, both of which are beyond my skill as a healer. We will have to take this computer to Rivendell.
No commentsResurrecting a Cube
I’ve been working on resurrecting a g4 cube that decided to graze off a cliff sometime after the update to 10.2.x. I’ve seen a couple of other people having similar problems, but no conclusive results from all the various mac hardware voodoo procedures. Most of the problems appear to be centered on open firmware.
- On boot, it dumps into Open Firmware
- It’s possible to continue out of it by typing “mac-boot”.
- If it’s set to boot to 9.x, it hangs with a gray screen.
- It won’t even boot 9.1 off of an iPod.
- If set to boot OSX, it will boot
- Pressing C on exit from open firmware allows booting off of the cd. (only had OSX install disks, not sure about 9.x)
- None of the power buttons work, so once the machine is shut down, you need to unplug, wait and replug to restart it.
- Open firmware is set auto-boot=true and boot-command=mac-boot, so I don’t see why it’s dropping into console.
I’ve tried pulling the pram battery, but as I didn’t have the correct replacement, I didn’t get to replace it. I am guessing that that will solve the power button issue. resetting PRAM didn’t help the open firmware at all.
Googling afterward, it appears that the correct sequence for resetting nvram is “reset-nvram \n set-defaults \n reset-all” from the Open Firmware console. I’m also thinking of trying the cuda button if I can find it with a torx screwdriver.
Apparently useful links:
symptom vs cure more
more
a possible success
cuda switch location a faq
A toy I want…
Last fall I tried to setup a portable net access system — bluetooth for the phone, computer and new pda. It wasn’t really successful, partly because of the nature of the phone backhaul (it was just like a very slow modem) and partly because it’s a pain to have to decide to connect every time.
What I’d really like is a cellphone that had both a good data back haul and a wifi card. I don’t need much power in the wifi card, so battery life shouldn’t be too much worse than with the current bluetooth. 100Kbps for the data would be ok, but Verizon’s 250/100 service would be even better. A little bit of routing software, and I’d have my own portable hotspot, open or closed, anywhere that I can get a decent cell signal. It’s possible to piece together bits of this right now, given a smartphone w/data and a SD or Compact Flash Wifi card. I’m not sure that the WindowsCE smartphones have the capability to do the routing. They certainly have the memory and processor power to pull it off.
I’m pretty sure that Tmobile would never go for this sort of thing, since it would directly cannibalize their wifi hotspot business. But a company like Verizon who’s getting a name for the fastest data channels might be able to take a bite out of the hotspot business with something like this.
No commentsCareless CSS
For some reason IE5.0/windows treats “display: relative” as “display: ignore and create a big white patch where your content would be in a real browser”.
This has now been fixed. We return you to your regularly styled drivel.
No commentsCareless
I couldn’t forgive him or like him, but I saw that what he had done was, to him, entirely justified. It was all very careless and confused. They were careless people, Tom and Daisy — they smashed up things and creatures and then retreated back into their money or their vast carelessness or whatever it was that kept them together, and let other people clean up the mess they have made. . . . (The Great Gatsby, F. Scott Fitzgerald)
I was amused and excited about the Moon and Mars missions that Bush announced. While it seemed a naked jobs ploy for Florida, Texas, and California, it did inspire. I’ve been reading a lot of SF lately, including the Mars Trilogy from Kim Stanley Robinson — that sort of mission grabs the imagination and takes it off the planet.
But the other side is that the manned missions have nearly no additional funding. So the Hubble Space Telescope, a source of some pretty amazing data over the last decade, will be abandoned within a few years as it’s systems degrade. There won’t be a replacement on the scene for another decade.
It seems careless to not maintain an instrument that has given us so much good data, when the technical, monetary and human resources exist to do that maintenance. But I’ve realized that it’s not so much carelessness as premeditated instrumenticide.
Everything that has a price that this administration proposes is a method of killing off something that they don’t support. Everything.
No commentsMore UML on Debian Stable
More stuff on usermode linux. (So I don’t forget). There’s a tun group to allow access to the tun device. The user starting the usermode linux needs to be part of that group. This all needs to be done on each boot, so it’s best if you add it to a script. Everything here needs to be done as root on the host machine.
/sbin/modprobe tun /bin/rm /dev/net/tun /bin/ln -s /dev/misc/net/tun /dev/net/tun /bin/chmod 660 /dev/misc/net/tun /bin/chgrp tun /dev/misc/net/tun /usr/sbin/tunctl -u erics /sbin/ifconfig tap0 192.168.20.1 up /sbin/route add -host 192.168.20.2 dev tap0 /usr/sbin/tunctl -u erics /sbin/ifconfig tap1 192.168.20.1 up /sbin/route add -host 192.168.20.3 dev tap1
And in the other unrelated thing, I’ve ruled out the drivers and the noapic flag on the netdev watchdog timeouts. I don’t think I saw these on my -bf series kernel, so I may try to go back to one of those.
1 commentCar Salesman
I’m not sure what it is about car salesmen, but they don’t seem to be living in the 21st century. Having visited 4 over the weekend, I have noted some similarities in their behavior.
- One of them had not sold the car we were looking at to a family member. (Two actually, but when the salesman had to defer to the Guy Who Could Talk To The Man, there was some random story about a nephew’s wife and a fridge. )
- Every single one of them told us something that was verifiably false, normally disparaging about the other cars we were looking at. Also generally something that could be fact checked on the net in about 2 minutes. (E.g., the japanese national speed limit is ~35 mph, regarding japanese car crash safety. Ford resale value is 50% after driving off the lot. This is the only car like it in 5 states. Sticker + 30% is a reasonable starting point for negotiations.)
- Must keep talking. It seems to be some rule that if they run out of things to say then they will lose the sale. One of the guys at least realized that we just wanted a little space to measure the car.
It will be interesting to see how this all turns out.
No commentsUsermode Linux on Debian Stable
It took a little bit of futzing to get UML working on Debian Stable hosting a newly installed Debian Stable image.
- The usermode tools and kernel image installed with debian stable are very out of date. You will need to update them.
- You need a relatively new uml kernel. I used this non modularized kernel
- I made an updated .deb package of the 09/03/2003 version of the utitilities and installed them.
- I wound up installing a new host kernel using devfs. I’m not sure if this made the difference, but since I needed to install it anyway for the latest security patch.
- I needed to fix the alias at /dev/net/tun to point to /dev/misc/net/tun rather than just misc/net/tun. I also needed to give the uml host user rw access on it.
- I wound up using xterms as the console.
- The startup command line I’m using is ./linux ubd0=server1 con=xterm con0=fd:0,fd:1 eth0=tuntap,tap0,,192.168.20.1
- When setting up the network connection with presetup TUN/TAP devices, you need to make sure that the device name is in the second field.
- When setting up firewall rules for your new virtual machine, don’t forget that you need to both enable forwarding and masquerade.
- If you install a debian system image using debootstrap, the second (normally after reboot) stage is /usr/sbin/base-config
Other Useful Links:
- http://trilldev.sourceforge.net/files/remotedeb.html
- Notes on creating a woody filesystem image
- http://uml.openconsultancy.com/
And a not related useful link: on netdev watchdog timeouts Update, didn’t help. Leading candidate now is some sort of apic bug or cheap network cards, since I have an old AMD/via system and um… cheap network cards. I may just write a script to down/up the interface when I see the note in the kernel log.
No comments