*scottstuff*

I’ve been slowly working my way through Neal Stephenson’s Quicksilver. It’d be a faster read if I actually had time to sit down and read it–as it is, I don’t think I’ve managed more then 30 minutes of uninterrupted reading in over a week. So far, it’s proving to be a fascinating read, but not for the reasons that I expected–the history in this book is really the fascinating part. I’m not sure how historically accurate all of it is (I should really check when I’m done), but the way he presents the origin of the scientific age is deeply thought-provoking.

I’ve been involved with the Net for years, starting in the fall of 1990 at the University of Chicago. I remember the first release of Mosaic clearly; since I had Motif for my Linux box I eagerly downloaded and compiled it, and then posted a Linux binary of it for download within a few days of its release. I remember days when I started at the “What’s New” page and stopped when I’d read the entire Net. I remember www.mtv.com, the first time. I remember the first round of “Internet” articles in the mainstream media, like The New York Times. I remember when the Puget Sound Computer User somehow screwed up and claimed that I’d written Mosiac :-).

As clear as anything, though, I remember the day that The Net became a part of The Normal World. I’d just walked out of the library and there it was–a shiny new UPS truck with “http://www.ups.com” on the side. Clearly, the Net Had Arrived if UPS thought it was worth it to print a URL on the side of their trucks.

Over the years, they seem to have dropped the “http://” from the front, as a kind of backhanded statement on how common web browsing has become. After all, http:// is almost completely redundant with modern browsers.

This morning, on the way to work, I saw another new, shiny UPS truck, with their newest design. With a bit of sadness, I noticed that the “www.ups.com” was gone from the side. I suppose it’s fitting that UPS trucks have come to symbolize the growth of the Net to me–from shock, to familiarity, to such complete ubiquity that there’s no reason to even print UPS’s URL on their trucks.

When I was young, I had a boxload of 78s with an assortment of kid’s music–American folk music, Sesame Street, and so forth. I particularly remember ”Turkey in the Straw” going around and around on my little plastic record player.

I discovered today that Cindy has Gabe hooked on “Turkey in the Straw,” except he doesn’t listen to it on 78s, or cassettes, or even CDs. Nope, Gabe’s modern: he uses MP3s to listen to his kid’s music. Sigh. I’m getting old.

I’m slowly re-organizing my online photo gallery in preparation for adding at least a few of the huge pile of pictures that I’ve taken over the past few years. The current gallery will be unmanageable if I just keep adding things. For now, I’ve just added one extra level of subdirectories, but I’m still not really happy with things. I’m using a slightly tweaked version of Album for producing the gallery; it’s better then most of what I’ve looked at, but it’s far from perfect. I suppose what I really want is a web-based version of iView MediaPro, but that’s kind of unlikely.

I’ve tried a bunch of gallery software, and I’m utterly incapable of coping with anything that’s designed for web-based uploading. It’s just too slow when you have 50 pictures that you want to add, much less the thousand or so that I’d like to have online. Also, D60 images are just too big for my poor little Athlon 700 to resize on the fly, so the gallery needs to have statically generated images. That rules out things like IMS that really want to generate images on the fly. I’d really like to be able to search and sort on IPTC and EXIF fields, and maybe even to be able to dynamically build new albums from existing images, but that’s beyond the current state of the art in free web galleries, and I don’t have time to write one right now.

We’re starting to get to the point at work where communication between developers is getting difficult. So, I’ve spent about half of today working on getting a wiki working, along with a weblog. I’d really like to have a single, integrated wikilog, but I can’t really a suitable one. So, a bit of hacking, and I have a wiki that does RSS and a web log that can link WikiWords to the wiki. It’s not perfect, but it’s good enough for now.

Now, all we need is content, and to train people to use it.

Cool. Legos come to the Seattle Center. I wonder if Gabe would be interested–he’s almost old enough for “real” legos now.

I’m taking the day off today. The last month and a half at work have been this huge grind working up our 1.0 release at the end of August, followed immediately by another two week project with 10 hour days and looming deadlines.

I need a break. So, I’m taking today (and hopefully tomorrow) off. Today, we’re going to the zoo. Gabe loves the zoo, it’s been all he’s been talking about for days. It’s not like I don’t enjoy it too; it’s a nice place to take pictures, and Gabe always enjoys looking at Zoo shots.

Plus, this may be the last stretch of good weather this summer, and it’d be a shame to waste it.

I’ve been meaning to put restaurant mini-reviews here for a while, but it’s been a few weeks since I’ve eaten at anyplace worthy of mention. I had lunch today at The Buffalo Deli in Belltown. I had a roast beef sub with everything, and it was probably the best sandwich I’ve had in years. Pretty much everything was perfect, except the price ($8.50 or so). The beef was great, the bread was great, the flavoring was was subtler then you’d expect on a sub sandwich, but it all just worked. Highly recommended.

One of the interesting bit of information coming out of IDF this week is that Intel is planning on making future x86 processors virtualizable. Since people have been asking for this since the days of Desqview/386, Intel doesn’t exactly get any promptness points for the announcement. Assuming that they doesn’t screw things up, the goal is to make it easier for future PCs to support multiple virtual servers, all running different OSes on the same hardware without needing complex pieces of software like VMWare. Mainframes have been doing this for decades, of course, and that’s one of the big selling points for running Linux on them–there have been demonstrated cases of people running over 40,000 virtual Linux servers on a single mainframe.

In typical Intel fashion, they’re giving it a obscurely trendy high-tech codename: “Vanderpool Technology.”

Presumably, Intel is aiming to use Linux as the “hypervisor” and then run some mix of Linux and Windows in virtual partitions. This meshes nicely with my “disposable virtual server” plans, so I can’t really wait to see exactly what they’re planning. I’m such a sucker for new toys.

If I remember correctly, the x86 is already mostly virtualizable; there are only a small handful of instructions that you can’t trap and can’t emulate easily. So, it shouldn’t be particularly difficult for Intel to add another CPU flag that’ll turn on traps for the remaining ops that need to be trapped. The problem is that only leaves us with a virtualized CPU; generating a complete virtual PC is a lot more work. I’m not sure what they’re planning on doing for a BIOS, storage, networking, graphics, and so on. Mainframes have spent a lot of time over the years designing hardware that works well in this sort of environment, but it’s a completely new area for PCs.

I’m not sure how this will mesh with User-mode Linux. UML is serving two purposes right now; it’s able to be used for (somewhat slow) virtual servers, but it’s also a version of the Linux kernel that’s easier to debug and upgrade then traditional Linux running natively on real hardware. I’m not sure how it’ll interact with better in-CPU virtualization.

They’re doing it again. Verizon DSL is basically down for me right now. It’s been spotty for days, dropping on and off the net for an hour or so at a stretch, but now it’s not even coming back up right:

# ping -s 1000 206.253.194.65
PING 206.253.194.65 (206.253.194.65) 1000(1028) bytes of data.
1008 bytes from 206.253.194.65: icmp_seq=5 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=8 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=9 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=12 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=13 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=14 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=18 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=27 ttl=247 time=125 ms
1008 bytes from 206.253.194.65: icmp_seq=35 ttl=247 time=125 ms

--- 206.253.194.65 ping statistics ---
38 packets transmitted, 9 received, 76% packet loss, time 37006ms
rtt min/avg/max/mdev = 125.389/125.490/125.641/0.246 ms

Lame. I really need to switch.

Okay, I’ll admit it–I lust after all of the shiny new handhelds that keep coming out. Sony’s PEG-UX50 looks amazing, and the Japan-only Zaurus SL-C760 is so far ahead of all of the other handhelds on the market that it’s scary. I mean, it has 4x the pixels of all of the Pocket PCs on the market, plus a (supposedly) usable keyboard.

There’s only one problem with all of these. I don’t have a clue what I’d do with them. I’ve owned a Palm III and a newer Clie, but they haven’t exactly changed my life, and in the end I stopped carrying them around, replacing them with a cell phone that syncs with my laptop. The phone is horrible for entering new information, but it can do it, and it’s okay at displaying calendar entries, so it does most of what I need on that front. Obviously, it’s pretty good at storing phone numbers, and iSync on the mac is actually usable now. In short, it all Just Works, which is what I want.

So, what would I do with a spiffy little handheld? Well, mostly play games. Store reference texts. Sometimes take notes. Maybe browse the web and deal with email. I mean, I’d dearly love a pocket-sized wireless web browser, but I’m too cynical to believe that any of them on the market really work right yet. I played with a Palm Tungsten C a few months ago, and the size, display, and keyboard were nice, but PalmOS really isn’t up to the Internet yet. It doesn’t multitask even slightly, and the startup time for the web browser was terrible–flipping back and forth between the browser and another palm app would drive you insane. Supposedly Sony’s browser is better, but I’ve never actually been able to get my hands on a Clie with a network card to play with.

The Zaurus obviously doesn’t suffer from PalmOS’s problems, since it runs Linux, but it has a host of other problems. Like, it’s completely incapable of syncing with my Mac. Plus, it isn’t easily available in the US. You need to tweak it to get it to display English at all. You need a CF wireless card to browse the net, and most of them stick out far enough to be a problem. Assuming that you want a wireless card, then you need to use a SD card for storage, but Linux’s SD support is weak. You’d be better off with a SD wireless card and a CF storage card, but the Zaurus doesn’t do SDIO, so the wireless card won’t work.

And so on. None of the Zaurus’s problem are really fatal by themselves, but the poor thing dies a death of 10,000 papercuts.

What I really want is for Apple to come out with a handheld, but they keep swearing up and down that it’s a dumb idea and they aren’t going to do it. I’d love to see a slightly bigger-then-average handheld with wireless and bluetooth, and maybe even a GSM phone, a VGA (or better) display, and tablet-style clamshell design. I’d really love to see it with a 1.8” hard drive, or even one of the cheap 1.5GB drives that are starting to show up in MP3 players, although that isn’t really big enough to displace my iPod. I want a camera, either a ~2MP fixed focal length camera for taking pictures of whiteboards and documents, or a lower-resolution camera for taking video. Or one camera that can do both, although it’s going to eat batteries. Basically, what I want is a handheld that I can use to listen to MP3s, make phone calls (cellular or VoIP) over a bluetooth headset, take notes during meetings (with pictures), write short documents, search through text, maybe play with an ebook or two, and play back ripped DVDs if the mood strikes me, all while browsing the web, dealing with email, and occasionally IM. All for under $700 or so, with good battery life.

Yeah, I know, I’m dreaming.

Why is DSL so hard to get? I mean, I live in the suburbs, and I’m well within DSL range, but actually getting it the way I want it seems to require a secret handshake or something that I don’t possess. What I want is easy: ADSL, at least 768 down and 256 up, a static address, and the ability to run servers. Not exactly rocket science. Speakeasy, for one, would love to sell it to me. Except they can’t. Verizon tells Covad that I’m living outside of ADSL range. Even though I have two COs within a mile of my house. Even though Verizon itself sells me ADSL. Even though my former neighborhood DSL installer claims that I’m an easy 5-6,000 feet to the CO south of my house, apparently Covad thinks I’m wired to the CO north of my house, so they can’t touch me.

I’ve been using Verizon DSL for over a year now, and they’re not as bad as they could be, but they’re an ILEC, so that’s not saying much. They’re routinely down for a couple hours at a time, they charge through the nose for static IP addresses (I could pay half as much for twice the speed, if I could cope with dynamic addresses and no servers). I used to have Speakeasy, because I believed them when they said all I could get was IDSL at $90/month for 144 kbps each way.

From the look of things, I could get Speakeasy to sell me 1.5/256 for $60 month if I could just get Verizon to fix their database, or at least explain why they can sell me DSL but Covad can’t, even though they’re both in the same CO.

It’s Network Cleaning Day. It’s been too long since I last cleaned up the computer room, much less the software lurking on random servers around the house. I’ve had one server that I’ve been unable to log into for over a month, and I just plugged a keyboard and mouse into it to fix it (it was a DNS problem; I kinda knew that).

The goal:

1. Add 4 160GB drives + 3ware RAID card to the file server. Actually get it serving files again :-).
2. Move DHCP server from file server to gateway box.
3. Install apt-proxy, so I don’t have to download the same packages over and over again for all 3 debian boxes.
4. Fix the slightly broken network connection on the file server. It loses packets and generates errors. Not enough to be really broken, just enough to occationally kill TCP performance. I’ve replaced the cable and switched switch ports in the past, so it’s time for a new NIC. This is actually a 3 year old problem, so it’s time to fix it…
5. Move the wireless network onto its own segment so I have more control over it.
6. Add a gig-e card to my file server, just so my laptop has a faster connection to the new RAID array.
7. Get rid of nasty excess cables.
8. Make sense of the mess of power strips.
9. Clean up so cables for my laptop don’t run down the front edge of the desk.
10. Get printing actually working right for once.

Should only take a couple hours, right?

More more quick storage note for the day…

Rik van Riel posted a nice list of research issues to the linux-kernel mailing list talking about open problems in computer science. One of the things that he points out is that, by some metrics, computers are actually getting a lot slower. For example, the amount of time it takes to read an entire hard drive, or the amount of time it takes to scan through all of RAM. The basic problem is that storage capacity is growing faster then storage speed can keep up. That throws things out of balance, which means that a lot of the old assumptions that we’ve grown accustomed to are starting to be wrong now, and will be really wrong in the future if things don’t change. For example, the optimal disk block size for reads and writes, focusing solely on throughput, is probably around 1 MB right now for most drives. Reading 1 MB only takes a bit longer then reading a single sector from the disk.

Interesting stuff to think about.

Anyone who’s been around computers for a while understands the “but what would you do with all that disk space?” question from friends and family that inevitably pops up every time you decide to add more drives. Around 18 months ago, I set up a ~14 TB IDE RAID array (spread across 8 boxes) at work, and actually had a bit of time to think about it. 1 TB of disk space is (roughly)

• 800 days of 128kbit audio (MP3, AAC, WMA, doesn’t really matter)
• 72 days of uncompressed CD-grade audio
• 8 days of uncompressed 96k/24bit 5-channel audio
• 40 days of 1 GB/hour TiVo basic-grade MPEG2 video
• 10 days of 4 GB/hour TiVo best-grade MPEG2 video
• 5 days of ATSC compressed HDTV
• 3.5 days of video from a DV video camera (they don’t compress well)
• 200 DVD-R disks
• 50 double-sided, double-layer DVDs
• 1 hour of uncompressed HDTV video
• 73,000 raw images from Canon 1Ds camera (15 MB each)

So, 1 TB is a lot of data, at least in terms of audio and individual videos, but it’s not very big compared to a lot of people’s DVD libraries, and it isn’t really enough to hold archives of broadcast HDTV MPEG streams. It really isn’t enough for editing raw HDTV.

Comparing this to the random bits of media sitting around my house, this tells me that I can probably use 2-3 TB today at home without any real problem, assuming that I’m interested in ripping DVDs and stuff off of the TiVos. If I upgraded to HDTV, then I’d probably need 10 TB or so to be happy. If I was interested in editing raw HDTV, then I’d probably want a lot more.

But, this also puts an upper bound on things. I mean, even 1 TB is enough for all of my normal audio needs. It’ll probably hold all of the music that I’ve ever listened to in compressed form. At 128kbps, you could listen 8 hours per day, 300 days per year for 8 years before you had to repeat. Even with raw CD audio, you’d make it most of a year. On the other hand, it’ll barely hold every episode of The Simpsons :-).

So, assuming that 19 Mbps HDTV is good enough for anyone, ever, and that I wanted a year of video (8 hours, 300 days), I’d need around 20 TB of disk space. That sounds fairly reasonable to me–I could have most of the episodes of most of my favorite shows, plus a ton of random stuff sitting around. Bump that up to 100 TB, and I’m storing stuff that I’ll never, ever watch. To phrase it the other way, that’s two channels of HDTV, around the clock, for 9 months. If I was really serious about editing HDTV, I’d probably want a few more TB, but odds are it’ll be compressed at 1/10 or so the compression ratio most of the time, so 10 TB would get me 50 days or so of editable video. That’s more then enough for me.

All in all, I don’t know what I’d do with over 100 TB. Until we get there, though (about 12 years, according to Moore’s Law), I’m pretty sure that I’ll be able to fill up all of the disk space I can afford.

Now, backups are a different issue…