Linux

Plopped my self down to check email and there's a weird one: "Hi Paul from Google". Of course, my first thought was 'Spam!' The subject is odd, a bit awkward, and spammers are known to put names in subject lines to sneak past the filters.

But, my filters are pretty tight. Why'd this one make it through I wonder?

Well, after checking the headers, I found out it was really from Google. Wow! So, I called the guy. He was really friendly and very clued in. Right off the bat, in the very first call, he asked me some very difficult questions. I could answer them, but trust me, they were difficult.

Two more really, really thorough technical interviews came next. Not much to say about them, but I can't over-emphasize that they are looking for quality engineers and are not going to waste time asking easy questions.

Wow! Got invited to NYC for a interview. I did say all along that this was my goal. I knew that thousands of people apply to Google ( I hadn't ) and that they only take the best of the best. What I did want, and eventually got, was a chance to visit the office, and run the gauntlet myself.

Took Amtrak up to the city, what a great way to travel. Found my way to the offices, and they're just like you read about. There's a central rec-room of sorts with air hockey, pool tables, massage chairs, etc. That day's TechTalk speaker was Seth MacFarlane, from the Family Guy. There was a electronic drum set with headphones for a little venting. There was a major Lego center with thousands of bricks and completed and in-progress projects laying around. There were also sweet little common areas stocked with snacks, drinks, coffee, the works. In all reality, it's neater than I thought it would be. It's the Wonka's of the IT world.

I was intellectually beaten and abused during the interviews. Let's put it this way: I know more than the average bear about Linux, UNIX and networking. The questions I had to deal with were masterfully devious. I got the distinct feeling that they were being good engineers and testing to failure. The only question was, did the test material fail before or after the requirement?

Well, I didn't make the cut. ( I did however get the shirt! ) Not sure how I feel about it. The good news is that I no longer have to face the decision on whether to move away from our families or not. I do love most parts about my current job too, so that's good.

The real benefit to this, other than a good story is that I'm re-inspired. I got the puddin' kicked out of me by intellectual big-leaguers. I've been working hard, but I haven't been really stretching myself to grow to a an entirely new place. That needs to happen now.

Thanks Google.

Ugh.. This fix sucked up enough of my time, that I decided to post on it. Maybe it will save someone else...

Ebay's HTML editor just would not show up in IceWeasel ( Firefox ) browser on a Debian Etch system. Killin' me for over an hour... Javascript, Java, Flash, then back to the Javascript console.

Then, finally, I stumble on Bug #408864, which tells me to change my User Agent string to Firefox.

It worked.

Just so you know...

For reasons I won't get into (and am not particularly happy about), I wanted to put up a windows partition on my workstation.

Windows doesn't want to install anywhere but your primary master, so I unplugged my hda (Debian) stuck the to-be windows drive in and installed Win2K.

Cool.

Get in, add a stanza to grub that looks like this:

title=Win2K
root (hd1,0)
makeactive
chainloader +1

Well, the passover to the windows bootloader chokes. I keep getting "Filesystem type unknown, partition type 0x7".

After a lot of pain and googleing, I figure out that the general problem is that when the windows install happened, it was the primary disk. Now that I put my linux disk back in everthing's not where it's' supposed to be (according to the windows bootloader.)

Solution: Right as grub kicks off on hda, switch the two drives so windows boot loader doesn't get confused:

title=Win2K
root (hd1,0)
map (hd1) (hd0)
map (hd0) (hd1)
makeactive
chainloader +1

Over the years, I've used a variety of bootable tools. Very handy stuff to have for building and repairing machines.

My tried and true for a long time was Tom's Root Boot, which was originally for a floppy. For a while, I was even burning that image to bootable CDs.

Time progressed though... The floppy is dead, and USB media and CD are fairly easy to boot. That brought about a whole series of really neat bootable projects. I moved on to the Ubuntu Live project. It has the tools I need (chown, grub/lilo, cfdisk), but it isn't exactly geared towards emergency work.

For some reason lately, I've stumbled across the Ultimate Boot CD. Wow. Reminds me of the old Norton Disk Doctor days. There are a boatload of utilities available here. All of the vanilla partitioning stuff you'd expect, but also:

• NTFS analysis and password recovery
  
• Hard drive utils for the various manufacturers
  
• Wiping utilities
  
• Burn-In applications
  
• Memory tests and identification
  
• Benchmarks
  
• BIOS Stuff
  

Overall, a real must-have. Too bad it doesn't fit on one of the smaller shirt-pocket size CDs... But beggars can't be choosers and this is one fine tool to have in the old emergency kit.

On my webhost, I've got an eight-day logrotation of my apache logs. I didn't set this up (and can't change it), so it is what it is...

Now, I run rsync every day to get those log files, MySQL dumps, all kinds of good stuff. There's a problem with this though: as logrotate does it's thing, each of the 1-8 files get clobbered in turn. So, my logs don't accumulate.

What I want to do is to concatenate the files into a local master log that I can analyze over time.

Ever want to tar a lot of files (> 12000) and hit this limit? It's not tar's fault, nor is it 'mv', 'cp', or whatever else. As a matter of fact, it's not really even from your shell. Each shell session has a pre-configured amount of storage with a hard limit. To check it, thry this:

getconf ARG_MAX ===> 131072

What you'll need to do is to work around this. For tar, I found some good advice on pre-building a list of file-names and then passing the one pre-saved list to tar. That worked fine:

# cat names of all desired files. redirect to one file
find . -iname '*.daf' > my_file.lst
# run tar, passing that list in
tar czvf my_archive_file_name.tar.gz --files-from filelist.txt --remove-files

Another thing you can do is to use 'xargs', which is documented in the GNU FAQ:

find htdocs -name '*.html' -print0 | xargs -0 chmod a+r

Oh, man is this nice. I've been strugling a while to figure out how to run CPAN as a user without root access. I finally came across a nice, clear set of instructions. Good enough to document here for future reference.

HOWTO: Run CPAN as a non-root user

I've been (happily) using OpenOffice for a couple of years now. The recent release 2 however uses a whole new document format. Of course, it can open older stuff. I've been manually upgrading formats as I need various documents.

That's a good way to get burned though. Ideally, you should figure out a way to batch convert any document formats that change. (Hence my affections for ASCII, but that's another post.) This article looks like it contains all the clue's you'd need to do that, but at this very moment, it looks like I've got other bridges to burn.

Here's a good article on mass-converting Office documents. It's written by a guy named Bob DuCharme, which is funny because I've had a document around here for ages in my reference directory written by him. It's a masterful quick and dirty introduction to operating systems, one system per chapter. In the work I do, you just never know what kind of system might present itself, so I've hung onto that article for a a long time.

(Funny how even on the Internet, it's still a small world.)

So, you have a linux box that's been running a while, and has had a lot of upgrades over the months or years. What happens is that you'll have support libraries that either have been superceded, or you'll have an application removed that was the last dependent package on a library.

If you want to clean those up, run this command (as root) until it comes up empty.

dpkg --purge `deborphan`

The first thing to happen is that debporphan runs. It returns a list of libraries that no package is dependent upon. The backticks (with the tilde, not the single-quote) are special. They take the results of the deborphan command and place those results into the dpkg command. In otherwords, dpkg gets a fresh list from deborphan each time this command runs.

Freeing one library, might create another library that can be purged, so it might take a couple of shots to be thourough.

Maybe you're not too web-savvy. Maybe you've heard of terms like RSS, Atom, or FeedBurner.

They're all ways that you can take advantage of emerging web standards. Almost every new site will let you link to it, with an RSS reader. That means that they're exporting the content in a predictable way that you can read in an application other than a standard web browser. (If you've been around, you'll realize that it's very similar to USENET newsreaders. But, if you know what USENET is, then you probably know about RSS too.)

So, set yourself free. Stop wasting time getting lost in one site after another. Stop checking to see if a favorite site has new content or not. Go get yourself an upgrade, and check out a better way to read....

This article on 'web feeds' at wikipedia will lead you to all you ever wanted to know.

PS: If you're on linux, take a peek at the Liferea RSS reader.

For a while, I was pretty impressed with the bookmark service offered by del.icio.us. The concept is great: leave your bookmarks on a webserver, where they can be tagged, organized and accessed from anywhere. They can even be accessed by others. Even that, I'm OK with. Unfortunately, the entire project has been purchased by Yahoo. That's not a bad thing in and of itself, but that leaves your bookmarks one step closer to being fodder for Mega-Margeting droids.

The upside to del.icio.us is that they aggregate links. That means they have a really neat page of popular links, organized by topic. In a way, it's neater than google because it's organized and edited by humans. However, for privacy reasons, I've taken my bookmarks out of their system.

Instead, I highly recommend the Insipid bookmarking application. It takes a little bit to install, but once you get that done, you've got all the advantages of del.icio.us, but you get to hold your own cards.

PS: Insipid is perfectly happy importing bookmarks exported from del.ico.us, so you have one less excuse...

Well, I wanted to try something better than my old system, which was CMSimple. Not that it's a bad project. It just has a few shortcomings, especially the license.

I jumped over to Open Source CMS. It's a really neat site for checking out all sorts of CMS, blogs, etc. The neatest thing about that site is that they have publicly-accessible demos of all of the systems so you can test drive it.

Very impressed by the package you see now. It's called Serendipity and it's very tight, very polished, and the install is smooth like buttah. Pretty surprising, since I've never heard of it.

So far... highly recommended.

If you have an .ISO, no need to burn a cd! This command takes your ISO, and uses the loopback device to mount it directly into your filesystem. Very cool.

mount /tmp/track-01.iso /mnt/tmp -t iso9660 -o loop=/dev/loop3

This is really slick if you want to dump a filesystem from one machine to another, especially if you have a Knoppix or Ubuntu live CD handy...

If you want the files expanded at the other end:

tar -czvf - | ssh user@store "cd /tmp/backup; tar -xzvf - "

If you want a tarball at the other end:

tar -czvfp - | ssh user@store "cat >> /tmp/backup/filename.tar.gz"
tar --exclude /proc -cpvf - / | ssh -l dd of=backup.tar.gz

To watch what's going on outbound:

tcpdump -n -i ppp0 outbound

To watch in and out:

tcpdump -n -i ppp0 outbound

Make sure to use '-n' to suppress name lookups, or you'll recursively watch yourself lookup incoming scans!