Would someone explain to a simpleton why there isn't a multiple redundent opensource backup system for using recordable media?
My understanding of the failure mode of CD's and DVD's is that they go bad a bit or a sector at a time, not all at once everywhere.
Given that, it should be possible to design a multiply redundent data scheme to make recovery possible.
My prof told a story of one of the early vacuum tube computers. A 'bit' was stored in a shoe box sized module that had 7 tubes in it. ANY 5 of the tubes could die and the device would still work. (Everything was drawer mounted Grad students with shopping carts full of vacuum tubes would run through the halls replacing tubes on the fly)
E.g. Reed Solomon codes can make it possible to correct an X bit error, and detect a Y bit error in a block. I don't know if these are the best encoding scheme for redundency.
Record onto two disks, both with RS error detection. If you know which block is bad, you have good odds to recover the corresponding block form the other disk.
Use a format so that losing one sector only loses you one sector of a file, not the entire file/directory/disk. This requires additional redundency for meta data, and it will mean don't use compression or encryption schemes that require the previous block to be read correctly to read the current block.
If someone does this I suggest a tier of standards, based on the desired probability of full recovery, and the probability of a read error on the media.
So for example Level 1 is based on making 2 copies + RS codes + redundant metadata on a single disk.
Level 2 is based on making at least 2 copies, and enough recovery code so that ANY one scratch across the face of the disk can't delete all the data.
Level 3 is based on making two copies of the disk, and labeling appropriately.
Level 4 is based on making 3 copies of the disk, with one labeled for cold storage.
It may be to do this, you will need to modify the media writer to access arbitrary locations on the media -- e.g. if the chunk of the disk that says what the disk is is bad.
As sysadmin at a school I assigned passwords to the kids, and set the system so that they couldn't change them. They could come to me or the school secretary to find out what it was as often as they wanted, but after the first request, asking me cost them 10 pushups.
Passwords were generated by creating a digraph tree from a large block of text (a mail spool file)
Pick a digraph at random. ac Now pick a digraph at random from all the ones starting with c and add the second letter. -> ck -> ack Repeat.
The effect of this is that while random, the distribution of vowels is such that about half of them are pronounceable. It's much easier to remember a random word (even with non alphabet characters in it) than total alphabet soup. E.g. fliP.nibblit3! vs qVrET2-Yp]
It's a smaller name space, but avoids much of the post-it security issue.
The rest of your points are relatively good, though the trees on the south side of the building will only help you in the summer, so only practical in temperate no-winter areas.
Endquote
At latitude 54 North (Edmonton Alberta) south side summer trees are still a good idea. In the winter enough sun comes through the trees to make a signficant difference.
Our house has half of it's window on the south side. We regularly get 10C over ambient during the day, if we leave windows shut. This means that the furnace doesn't run for three months of our heating season during the day. Much of this time we build a single fire in the evening as the house cools.
If anything, here the issue is the opposite. Our air conditioning need is small. (We average under 3 weeks a year with temperatures above 80F) With windows open on both floors, the chimney effect keeps the house cool.
The winter heating advantage is big enough that you don't really want 40% of the light blocked by twigs.
For any form of online news service to get my dollar it has to be done very differently:
A. Editorial content:
Newspapers focus on noise. Most of the articles as a story develops are irrelevant detail or 'juicy' stuff that panders to our taste for death and sex.
During the fiasco in what was Yugoslavia, the American media portrayed the Slavic faction as evil oppressors. I happened to catch an episode on CBC's "Ideas" that went into the history of the region. It's a lot more complex than I had thought. I'm training myself to say, "I don't know enough to have an opinion." on any issue that I learn from the media.
To get my buck, electronic papers need to be linked to the back story -- how the hell did we get into this mess.
Another feature that the media could do in general would be to follow up last year's stories.
Remember the amount of coverage the Exxon Valdez disaster got while it was happening? (Young pups: Big oil spill off the Alaska coast) Did you see any coverage of the area two years later? What's the impact of Chernobyl? How has Katrina changed policy in FEMA
I want to be able to tell my epaper that, "I'm not interested in stories about encounters between sports teams, but yes I am interested in sports as a cultural element."
Or that I want any story that takes place within 15 blocks of my house, or in any of the schools my kids are in. Much of the problem with a big city newspaper is that it too much of everything, and it's hard to sort through the cruft.
I want to be able to tag a story, "Archive this story and it's follow-ups, add these personal tags to the list."
I want to use my finger tip as a highlighter.
B. Advertising content: I could buy a newspaper just for the ads and the comics. If the ads were done right:
1. The ad is relevant either to me or to the story. 2. I can train the ad engine by marking one ad per screen as interesting or obnoxious. 3. The ad explains what is being offered, and who is offering it. No blind phone numbers. 4. The ad links to either the newspapers web site or the advertisers web site contained within the newspapers frame. (when the paper is online) or to a reduced canned copy. (when offline) The frame would contain elements allowing people to fuss about errors in item 6. 5. Ads too can be highlighted and saved. 6. The newspaper requires all ads to have links to third party reviews of the product. Part of the newspaper's ad department is to check this aspect for complete and representative content. E.g. Electronics would not only have links to the appropriate section of Electtonics World but also to the right page on epinion.com. 7. Advertisers were required to be truthful. (Yes there is an advertiser's code, but there's a lot of deception still in what they neglect to say.)
I am a subscriber to the Economist online. They still don't have it right. Stories are sound bite sized with no links to back stories, no explanations or too much simplified explanations for events and technology.
No real world-impact (or even state impact) event can be explained from scratch in 500 words. The world isn't that simple.
But the other option is that you serially create evidence (including, but not limited to video) that indicates that *you* did something. Cry wolf a dozen times this way, then do something for real with some of the same style of created evidence.
I can see a whole genre for high school and college students to try to create videos that cause the police to run around in circles.
As to shots of the police breaking their own laws: Those get uploaded to youtube, or perhaps someone will create gedankenpolitzie.org (thought police) as a host site for such vids. (Maybe wikileaks would host it...)
Yesterday my meter was replaced. The old mechanical one replaced with a new electronic one. Not only do they no longer have to send a meter reader around every two months, but they (and I) will be able to pull down hour by hour readings.
Power companies also have to plan for the peak time. One of the things they are trying in places is variable rate power. Peak power costs more than off peak, which in turn is more than night power. In California it's worth while to run the air conditioners at night, and cool a honking big tank of water. Then during the day circulate the cold water.
I chuckle at everyone's whining about non-capped bandwidth.
I pay $60/month for theoretical 2 Mbit down, 500 Mbit up. (Anik satellite link. )
I say theoretical as:
* 80% of that is the max throughput, what with the message overhead.
* From after school to 9 p.m. it may be 1/3 of that, due to everyone and his dog using the system.
* Any one connection runs through some kind of bucket scheme, so if I want to download an ISO image the first 24 MB come down fairly fast, but after that it is only about 12 MB/hour.
While, as a previous poster remarked, everyone spent all day pulling youtube, the company would make the same money is true, people would find that the download speed would be disgustingly slow. At some point upstream there is a link whose capacity was planned for the average evening peak with it's mix of Youtube, gmail, Craigslist, and World of Warcraft.
You plan capacity knowing that. Usage peaks in late afternoon, early evening. Amoung the users, most of them are doing casual surfing, email, online games, that don't put the load that shared vidio, massive software downloads put on the system.
Even if the last link to the house is rated at 50 MByte/s, the aggregation into the trunks is not the sum of the leaf nodes.
You see this on local networks too. A classroom network connected to the server with a 100Mbit LAN is fine -- until class change time when 30 kids save their profiles at the same time. Each computer's link to the switch is fast enough, but the link from switch to server isn't. Or the server disk isn't fast enough.
In a non-monopolistic system, a company can best use their infrastructure by helping their users spread their use.
1. A widget in their browser toolbar that says what the current rate is.
2. A download widget/bit torrent client that can be configured to do it's work when the network is loafing, perhaps by checking a server for overall congestion.
So many ancillary services use petro fuels that even solar power has a carbon footprint.
A better way to look at it is Pounds CO2/kWHr.
I live south of the mine that provides Epcor's Genesee Generating Station with it's coal. (The new permit boundary is 800 yards from my house. It's going to be interesting having a dragline as a neighbour.)
Their newest project is to use a coal gassification technology, burn the syngas in a turbine, use the exhaust gas from the turbine to run a boiler. The CO2 exhaust from the boiler is separated out, compressed to a liquid, and is to be piped 25 km to an oil field for enhanced recovery. Later, when the field is depleted, this same area has deep saline aquifers (>2 km) where they can inject the CO2. At these depths, the CO2 shows little tendency to come out of solution, and eventually forms carbonates with the local rock matrix.
The net result is 85% less CO2 per pound of coal burned. Due to inefficiencies of the process, it's about 75% less CO2 total. (Takes energy to compress the CO2.)
Coal gassification is a good step.
Another technology on the horizon is solid carbon fuel cells. They promise an efficiency of 70-80% and deliver a nearly pure stream of CO2 for sequesterization. The higher efficiency and cheaper separation technology combine to cut the CO2 a bunch further.
Clean coal is probably the best answer we have, at least until these ultra capacitors become main stream, and allow the use of intermittent power sources for continuous loads.
Buy 4 used computers. NOT dells. Not to disparage dells, but they have non-standard form figure power supplies.
Two machiens replace his current machines.
The other two are parts.
Put a second drive in each running machine. Doesn't have to be big, but disks that sit, stick.
Put UPS or at least decent surge suppressors on the working boxes.
I ran a computuer lab in a rural school for 5 years. With 60 machines I lost 3-4 power supplies per year, and 4-5 monitors per year. Lost 5 monitors in 3 days. Surge.
Adding an extra cabinet fan set to vent through 3 removed expansion slot covers is a good idea.
2. Software. If you are an engineer you likely need to learn Autocad, or a raft of simulation packages. If you are in any data intense environment, you need to use stats packages. If you use math you may want mathematica, maple. If you are an arts major you may need access to Photoshop, adobe illustrator, CAD. If you are a comptuer sci major you may need access to multiple OSs.
Some of this can be handled with site licensing. Much of this can be handled with Virtual machines running on hefty hardware, with the student's laptop being a terminal.
The future of computer labs may be:
2-3 computers for use by students who don't have a laptop, just want to read their email, have to send a file as a genuine Word 2007 document.
The rest of the room are tables with an internet connection. (Bring your own cable.)
At the back there is a printer.
In addition there may be special purpose labs: E.g. E.E. labs where every station has an EEPROM programmer. Media labs where every station has three monitors, and Final Cut Pro.
Compute labs where the machines are simple front ends for working with clusters of computers.
I too could drop my power bill by 30 bucks a year. But I won't do it even for $100 per year.
Reason: State: At present I have 4 browser windows open, each with a dozen tabs, about 40 terminal windows open eitehr ready to run a command or running vi.
I have digikam running on one virtual desktop, XP running under virtualbox in another desktop.
If I shut down my box I figure it takes me about an hour to get back to where I was.
Given that I do something on my computer 7 days a week, spread out over a 16 hour day, it's not worth it. (I've experimetned with hibernation. Doesn't work well with this hardware. It hibernates beautifully. And never wakes up.
I do let my dual 21" CRT monitors go to sleep.
My wife knows someone at work who comes in, turns her computer on, and goes for a cup of coffee, chats with a co-worker, touches base with her boss, checks her mail box for mail, all waiting while her computer boots and gets all it's ducks in a row. Then she comes in and logs on. And waits another 3 minutes.
Lets take the 36 buck figure at face value. Hell lets call it 240 dollars. -- 20 bucks a month.
The average work month has 20 work days in it. A buck a day. Here the average computer user is likely to get at least $20 per hour. Sure receptionists get 15, but engineers get 40.
A dollar is 3 minutes of employee time.
If we use that 36 buck figure we are at 13 cents per day. Something like 24 seconds.
So tell me again how this saves money?
(Now if you got users to set their bios to start their computer 10 minutes before they arrived....)
Coal plants alter their power levels all the time.
They just can't do it fast. Typically they make about 2% more steam than they need. This handles the transient spikes. As demand picks up they increase the feed rate for the powdered coal, this boils water faster, requiring more feed water. Steam goes to turbine at faster rate, putting more torque on generator shaft to compensate for the increased emf back force from the increased demand.
when the plant is not used at capacity, the parasitic losses (power to keep the plant itself running) are a larger fraction of the total. Less efficient.
When the plant is used at over capacity, you have higher pressure losses in the turbine, hotter steam coming out. Less efficient. But the power curve has a pretty broad top.
Seasonally they put units on and off line.
They can also put a unit in idle mode. It makes steam, it spins the turbine, but there is no load on the turbine, so it takes very little steam. But everything is hot. I think from an idle to full power is only about 2 hours
We have a 10 MW hydro plant on our grid. It generally only releases water during 2 hours during the day, all clustered in the times that power generation is spiking, and most of that in 20 to 30 second spikes.
Unlike a power plant which takes about 48 hours to go from cold to full production, a dam can do this in seconds. When I did the tour there, I could watch the gate actuators for the squirrel cage turbine. They were constantly moving, several times per minute between about 1/4 power and full power. By using the dam this way, they were able to reduce that 2% just in case over production of steam to some smaller number.
The VAST majority of Carbon is in carbonate rocks.
While the amount of carbon is fixed, the fraction of it that is CO2 is not fixed. The fraction of CO2 that is in the atmosphere is not fixed.
There are various cycles that work on different time scales.
In a temperate deciduous forest there is a day cycle: leaf absorbes CO2 during the day converting it into sugar. Releases CO2 during the night as chemical reactions continue. The surplus is turned to leaf and exported to the rest of the tree.
year cycle: Tree converts CO2 + water into leaf. Leaf falls off tree and rots, becomes worm food, fungi food.
century cycle: tree grows up puts on mass. Dies, falls.
ocean cycle: CO2 is stored in surface water surface water is mixed deeper by wind and current action. In this way the top 400 meters of ocean is at equibrium with the atmosphere with about a 60 year time constant.
ocean deep cycle: Water mixing to deeper levels seems to take longer. I think I read an estimate of around 12,000 years.
rock cycle: Carbonate rocks subducted at edge of plates. Magma heat diassociates carbonates. CO2 released in volcanoes. CO2 taken up by algae, fall to bottom of sea. Time and presure form carbonate rocks. Residence time millions of years.
So long term solutions lie in understanding the residence time of these various cycles, and giving them a nudge toward the pools that make it a less immediate problem.
Conifers are darker than grass. Plant large areas to conifers, and you lower the albedo of the planet. Which increases the temperature.
Increasing the temperature at arctic latitudes, allows the tree line to move north, taking over relatively light areas of gravel and peat bog and turning them into dark forests.
This increases global warming...
Putting wood and paper in land fills actaully does do a pretty good job of sequestering it. Most landfills are anerobic, and while there is some methane production, most of the carbon just sits there. Saw a story about 50 year old newspapers pulled from an LA landfill still quite readable.
An even better way to sequester carbon is to turn it into charcoal, and mix it into the soil. Charcoal has a very long residence time in the soil, and acts as a sponge for nutrients, reducing leaching.
I have a Pogo Linux workstation. (Quiet!) with a dual core chip and 8 GB ram.
I run VirtualBox with a copy of WinXP for those applications that there isn't a good linux implementation for. (Sorry -- Gimp != Photoshop Glom != access) It's allocated 1.5 GB ram, and mounts my home directory from the real machine via Samba. I get acceptable performance on everything I do on it.
For test bed type applications, I think that RAM is your biggest consideration. You don't want the VM to swap to virtual memory. (VMVM? VM2?)
My brother tells a story of a bio grad student who was hired to teach biology for the last half year when the original one quit.
Because he didn't have a teaching degree, they moved all the 'real' students to to the other teacher, and he got all the students who 'needed a science to graduate before they go work in the mill'.
Most of these kids took shop too.
So he walks into his first class after spending all night going through dumpsters in town, gives each kid a large tin can and a mouse trap, with the assignment, "Build a live trap."
They spend the rest of the period brainstorming approaches.
Within a few days he has many live traps.
"What do mice like to eat"
Many answers.
"How could we find out"
Much discussion, but it boils down to, try an equal number with various food and see what you get.
So they do that.
Then he starts looking at where the traps were. And this leads into doubt about the first experiment. Some guys put their traps in the open, others by the wall. Some on the school property, some in their back yard.
More discussion.
More experiments.
Then he asks for what results they got last week. Nobody is writting stuff down. This turns into the importance of record keeping.
Then he asks, " Are you catching the same mouse over and over, or are you catching different mice?"
Whole can of worms about identification. They figure out a way of marking small diameter aluminum tubing and gently crimping it as a bracelet. Now they can tell if a mouse has been caught before.
"How big an area does a mouse forage?"
So they took over wild section of land behind the school, and placed the traps in a fairly even distribution over several acres. Most of the kids had between 6 and 12 traps now, and were visiting them twice a day.
Each kid recorded every mouse they caught, adding a new band when an untagged mouse fell into their clutches.
By this time the weather started getting nasty, as fall gave way to winter.
So the last few weeks to Christmas were spent analyizing the data. So he taught them about averages and means, and standard deviations, and regression testing. (This before calculators. All hand work.)
Remember this is a class that 'can't learn much, we can only hope they will be good with their hands'
Just before Christmas, the school imspector came in to observe. He looked at the neat notebooks, saw the earnest work the kids were doing, had to ask the teacher what a kid was talking about when mentioning mortality tables, and mean forage diameter differential between male and female mice.
The grad student explained.
The school inspector said that he wasn't following the curriculum, spoke to the principal, and had the grad student fired.
Open DNS is wonderful. I used it at the school where I worked. It didn't stop the porn, but it cut the incidence of accidental porn to nothing.
Also: Set google to "moderately safe"
The other point about having the computer in a public space is good. Doesn't have to be in the living room next to the noisy TV, but in a corner where people walk by on a frequent basis is good. And don't then use YOUR computer in a private space. Set an example.
I also think there is merit in doing homework on the dining room table. Folks are around to help. 'Room time' happens after 'chores' such as washing the dishes, and homework are done.
Still, they are going to run into content that is disturbing -- either new ideas, (Muslim family law -- role of women, sexual freedom) violence, bigotry, hate, and in general man's inhumanity to man. Start early and talk about what she's seen. Get her in the habit of asking questions to you. These discussions are springboards for talking about ethics and morality in general.
You have to decide about social networking sites. Much of that use depends on the crowd she falls into.
Remember too, that even if you control her access she will have less restricted access at friends' houses, at school, or even the public library. Teach them responsible use. Teach them to be skeptical, to figure out ways to verify, to look for consistency. Done this way you have a huge impact on her development as a responsible adult.
Somewhere around 10 MP you get to the limits of resolution for the typical reduced frame (2/3 the size of a 35mm frme) digital camera.
By extrapolation 16 MP would be around the limit for a full frame camera. (And then you have to buy a new set of lenses)
If you aren't pro you likely don't maintain your lenses clean enough to reach these resolutions. Indeed few lenses reach anything like theoretical resolution (Raleigh criteria)
I've got a Nikon DX70 which I love, but...
* It's performance in low light is noisy. * It only accomodates about 6, maybe 7 stops of illumination variation in a scene. Reminds me of shooting ektochrome slide film. Expose for the highlights, or for the shadows. You can't have both. Unless you use flash to fill in. By comparison film accomodates about 9 stops, and can be tweaked to 10-12 in the darkroom. (Much of the Zone system of the black and white photographers is based on these tweaks.) 10 stops is typical for a scene on a cloudless day.
The future:
With fast enough data collection you dump the CCD at 1000 frames per second, then post process for both noise reduction and image stabilization. Indeed: Collect during the time that the diaphram is closing and you may be able to get advantages of the speed of a wide open lens and the depth of field of a stopped down lens.
I suspect to do this you would have to build the processing into the chip. One proccessor per NxN pixel square? Each processor hands data back and forth to it's neighbors for handling image stabilization?
Set your camera up on a stand looking at a grey card. The camera takes a series of frames to calibrate the individual sensitivity of each sensor.
I have a version of the 5.56 beta released for Linux that I hacked to remove the time limitation.
It is *still* the best way to write any document that's more than 5 pages long and using more than 4 styles.
The equation edition is not as fast as latex, but much faster than word's equation editor. Results are between the quality of word and the quality of latex too.
As to *TeX. All is well if you want a standard document on a standard page. Just try to be creative with layout. E.g. A 20 page document that has 18 figures, and your style guide is that figures are either at the top or bottom of the page, but if there are two figures on facing pages, preference is that one is on top of the column, and one on the bottom, and that if the figure is on the top of a column, there should be a.5 pt line, 75% of the column width between the caption and the main text.
In a former life as a teacher, I went to the headmaster and said that I wanted to teach the grade 10 general science course differently.
I wanted to walk in the first day and talk about 4 elements, earth, air, fire and water. Then for a day show how this explains everything.
I wanted to teach about caloric and phlogiston and luminiferous ether, and with each show the explaining power they had, then the problems they ran into.
Ptolomaic astronomy with circles and epicyccles and eccentrics.
Geology would be fun, especially the bits with continental drift and catastrophism vs uniformism.
The goal, I explained to my head: Teach that science is a process, and not a body of knowledge.
If you want to run the Big 4 (word processing, spreadsheet, presentation, internet browsing) then FOSS solutions are idea. In my last school however:
* Staff used a client server commercial package called RenWeb for record keeping, lesson plans, and parent communication. Client was winsooze only.NET application with a raft of custom DLLs
* As the photography instructor I used and taught photoshop. I'm sorry but the Gimp just doesn't cut it.
* As an outdoor program coordinator I used Mapmaker Pro and Oziexplorer to build maps and to move data to/from my GPS. I've found no equivalent of Mapmaker that is OSS.
* As part of support we had a number of MS access quick and dirty database apps, each with a bunch of entry forms, and reports. I've spent days searching for an equivalent package that allows rapid database application prototyping. (I looked at kexi, knoda, rekall, bond, glom, pfm, and PgAdmin)
This doesn't mean that FOSS is not usable, but in most systems it will need to be a parallel system to paid software.
IF I were in charge of a school district I would look at doing it this way:
1. I would deploy student machines as being some form of terminal only. Students with laptops could use windows rdesktop, or VNC.
2. I would buy high end boxes stuffed with memory and run a bunch of instances of virtualbox or equivalent vmware, or remote X. Using immutable images in virtualbox makes for systems that are hard for the kids to corrupt.
3. Users on startup could select if they needed a windows session or a linux session or a mixed session. Tweak the system so that equivalent applications on linux ran faster.
4. Servers would run FOSS. Working around Active Domain isn't that difficult, especially if the vast majority of your windows installations are virtual.
5. In a few cases you need individual high powered machines. (Running photoshop in virtualbox is less than satisfactory)
I would make the transition as follows:
A. Using whatever machines I could get, I would start setting up ancillary servers. DHCP, YP, SMTP, POP, IMAP, NFS, SMB. This gives my admins experience in working with open source operating systems, but in a way that is easy to back out of. E.g. Initial Samba services can be for doing backups.
B. When this works, and admins are comfortable with it, I'd bring in 1 high power box per site, and set it up as a VB server. Initially it would serve only windows, and it would serve to visiting laptops. This is sold as a security measure to protect your network from laptops with unknown software.
C. Once this works, I'd convert one lab at each site to thin client setup. Their existing hard drives would be untouched while learning how the system works. Eventually this is sold as a cost cutting measure, as it permits running the lab machines for more years. After the admins are happy with the results, rip out the disks. This makes the lab both quieter and cooler.
D. Introduce VBox with saved state to staff. Being able to shut down in their classroom, and reopen at home and have the same machine state for their record keeping, lesson plans and so on will be a win. Or they can leave the machine at work, and have a client session from home.
E. With the money saved from not having to upgrade all the labs, I'd make a few high end machine sets for applications that are demanding such as Photoshop, and Final Cut Pro.
Depends on your environment. When I worked for a university department, I had a chronic problem with grad students hooking up all sorts of things to the ethernet ports. And with cheap ethernet devices this is a security concern.
We finally got switches that could be told, "This port belongs to this MAC" and outlets that weren't in use were not connected between the patch panel and the switch.
All this meant was that the grads had to learn how to clone the MAC address of the device they were replacing. Then nmap would tell me that the Linux box had turned into a Windows box.
Nowadays I expect that they would put in their own $40 wireless AP. Once a dozen of them did this, no one would be able to do anything on the air.matchedmatchedmatchedmatchedmatchedmatchedmatchedmatched
At one point I was in the middle of a transition between 10Base2 and 100BaseT networking. All done on a shoestring budget, with no permissible downtime. My wife bought me a Dilbert T-shirt. Dogbert is walking out of the server room, the room a shambles. He's carrying a baseball bat. "The Network is down, but I'm feeling much better."
I wore that shirt to work several times over the next few weeks. People gave me a bit more space and less hassle about the constant, "Excuse me while I run 11 cables above your office ceiling" interuptions.
Slashdot Mirror
Would someone explain to a simpleton why there isn't a multiple redundent opensource backup system for using recordable media?
My understanding of the failure mode of CD's and DVD's is that they go bad a bit or a sector at a time, not all at once everywhere.
Given that, it should be possible to design a multiply redundent data scheme to make recovery possible.
My prof told a story of one of the early vacuum tube computers. A 'bit' was stored in a shoe box sized module that had 7 tubes in it. ANY 5 of the tubes could die and the device would still work. (Everything was drawer mounted Grad students with shopping carts full of vacuum tubes would run through the halls replacing tubes on the fly)
E.g. Reed Solomon codes can make it possible to correct an X bit error, and detect a Y bit error in a block. I don't know if these are the best encoding scheme for redundency.
Record onto two disks, both with RS error detection. If you know which block is bad, you have good odds to recover the corresponding block form the other disk.
Use a format so that losing one sector only loses you one sector of a file, not the entire file/directory/disk. This requires additional redundency for meta data, and it will mean don't use compression or encryption schemes that require the previous block to be read correctly to read the current block.
If someone does this I suggest a tier of standards, based on the desired probability of full recovery, and the probability of a read error on the media.
So for example Level 1 is based on making 2 copies + RS codes + redundant metadata on a single disk.
Level 2 is based on making at least 2 copies, and enough recovery code so that ANY one scratch across the face of the disk can't delete all the data.
Level 3 is based on making two copies of the disk, and labeling appropriately.
Level 4 is based on making 3 copies of the disk, with one labeled for cold storage.
It may be to do this, you will need to modify the media writer to access arbitrary locations on the media -- e.g. if the chunk of the disk that says what the disk is is bad.
So why isn't this done?
Or is it, and I've just not heard of it.
As sysadmin at a school I assigned passwords to the kids, and set the system so that they couldn't change them. They could come to me or the school secretary to find out what it was as often as they wanted, but after the first request, asking me cost them 10 pushups.
Passwords were generated by creating a digraph tree from a large block of text (a mail spool file)
Pick a digraph at random. ac
Now pick a digraph at random from all the ones starting with c
and add the second letter. -> ck -> ack
Repeat.
The effect of this is that while random, the distribution of vowels is such that about half of them are pronounceable. It's much easier to remember a random word (even with non alphabet characters in it) than total alphabet soup. E.g. fliP.nibblit3! vs qVrET2-Yp]
It's a smaller name space, but avoids much of the post-it security issue.
Quote
The rest of your points are relatively good, though the trees on the south side of the building will only help you in the summer, so only practical in temperate no-winter areas.
Endquote
At latitude 54 North (Edmonton Alberta) south side summer trees are still a good idea. In the winter enough sun comes through the trees to make a signficant difference.
Our house has half of it's window on the south side. We
regularly get 10C over ambient during the day, if we leave windows shut. This means that the furnace doesn't run for three months of our heating season during the day. Much of this time we build a single fire in the evening as the house cools.
If anything, here the issue is the opposite. Our air conditioning need is small. (We average under 3 weeks a year with temperatures above 80F) With windows open on both floors, the chimney effect keeps the house cool.
The winter heating advantage is big enough that you don't really want 40% of the light blocked by twigs.
And newspapers make good garden mulch too.
For any form of online news service to get my dollar it has to be done very differently:
A. Editorial content:
Newspapers focus on noise. Most of the articles as a story develops are irrelevant detail or 'juicy' stuff that panders to our taste for death and sex.
During the fiasco in what was Yugoslavia, the American media portrayed the Slavic faction as evil oppressors. I happened to catch an episode on CBC's "Ideas" that went into the history of the region. It's a lot more complex than I had thought. I'm training myself to say, "I don't know enough to
have an opinion." on any issue that I learn from the media.
To get my buck, electronic papers need to be linked to the back story -- how the hell did we get into this mess.
Another feature that the media could do in general would be to follow up last year's stories.
Remember the amount of coverage the Exxon Valdez disaster got while it was happening? (Young pups: Big oil spill off the Alaska coast) Did you see any coverage of the area two years later? What's the impact of Chernobyl? How has Katrina changed policy in FEMA
I want to be able to tell my epaper that, "I'm not interested in stories about encounters between sports teams, but yes I
am interested in sports as a cultural element."
Or that I want any story that takes place within 15 blocks of my house, or in any of the schools my kids are in. Much of the problem with a big city newspaper is that it too much of everything, and it's hard to sort through the cruft.
I want to be able to tag a story, "Archive this story and it's follow-ups, add these personal tags to the list."
I want to use my finger tip as a highlighter.
B. Advertising content:
I could buy a newspaper just for the ads and the comics.
If the ads were done right:
1. The ad is relevant either to me or to the story.
2. I can train the ad engine by marking one ad per screen as interesting or obnoxious.
3. The ad explains what is being offered, and who is offering it. No blind phone numbers.
4. The ad links to either the newspapers web site or the advertisers web site contained within the newspapers frame. (when the paper is online) or to a reduced canned copy. (when offline) The frame would contain elements allowing people to fuss about errors in item 6.
5. Ads too can be highlighted and saved.
6. The newspaper requires all ads to have links to third party reviews of the product. Part of the newspaper's ad
department is to check this aspect for complete and representative content. E.g. Electronics would not only have links to the appropriate section of Electtonics World but also to the right page on epinion.com.
7. Advertisers were required to be truthful. (Yes there is
an advertiser's code, but there's a lot of deception still in
what they neglect to say.)
I am a subscriber to the Economist online. They still don't have it right. Stories are sound bite sized with no links to back stories, no explanations or too much simplified explanations for events and technology.
No real world-impact (or even state impact) event can be explained from scratch in 500 words. The world isn't that simple.
But the other option is that you serially create evidence (including, but not limited to video) that indicates that *you* did something. Cry wolf a dozen times this way, then do something for real with some of the same style of created evidence.
I can see a whole genre for high school and college students to try to create videos that cause the police to run around in circles.
As to shots of the police breaking their own laws: Those get uploaded to youtube, or perhaps someone will create gedankenpolitzie.org (thought police) as a host site for such vids. (Maybe wikileaks would host it...)
Yesterday my meter was replaced. The old mechanical one replaced with a new electronic one. Not only do they no longer have to send a meter reader around every two months, but they (and I) will be able to pull down hour by hour readings.
Power companies also have to plan for the peak time. One of the things they are trying in places is variable rate power. Peak power costs more than off peak, which in turn is more than night power. In California it's worth while to run the air conditioners at night, and cool a honking big tank of water. Then during the day circulate the cold water.
I chuckle at everyone's whining about non-capped bandwidth.
I pay $60/month for theoretical 2 Mbit down, 500 Mbit up.
(Anik satellite link. )
I say theoretical as:
* 80% of that is the max throughput, what with the message overhead.
* From after school to 9 p.m. it may be 1/3 of that, due to
everyone and his dog using the system.
* Any one connection runs through some kind of bucket scheme, so if I want to download an ISO image the first 24 MB come down fairly fast, but after that it is only about 12 MB/hour.
While, as a previous poster remarked, everyone spent all day pulling youtube, the company would make the same money is true, people would find that the download speed would be disgustingly slow. At some point upstream there is a link whose capacity was planned for the average evening peak with it's mix of Youtube, gmail, Craigslist, and World of Warcraft.
You plan capacity knowing that. Usage peaks in late afternoon, early evening. Amoung the users, most of them are
doing casual surfing, email, online games, that don't put the load that shared vidio, massive software downloads put on the system.
Even if the last link to the house is rated at 50 MByte/s,
the aggregation into the trunks is not the sum of the leaf
nodes.
You see this on local networks too. A classroom network connected to the server with a 100Mbit LAN is fine -- until class change time when 30 kids save their profiles at the same time. Each computer's link to the switch is fast enough, but the link from switch to server isn't. Or the server disk isn't fast enough.
In a non-monopolistic system, a company can best use their infrastructure by helping their users spread their use.
1. A widget in their browser toolbar that says what the current rate is.
2. A download widget/bit torrent client that can be configured to do it's work when the network is loafing, perhaps by checking a server for overall congestion.
So many ancillary services use petro fuels that even solar power has a carbon footprint.
A better way to look at it is Pounds CO2/kWHr.
I live south of the mine that provides Epcor's Genesee Generating Station with it's coal. (The new permit boundary is 800 yards from my house. It's going to be interesting having a dragline as a neighbour.)
Their newest project is to use a coal gassification technology, burn the syngas in a turbine, use the exhaust gas from the turbine to run a boiler.
The CO2 exhaust from the boiler is separated out, compressed to a liquid, and is to be piped 25 km to an oil field for enhanced recovery. Later, when
the field is depleted, this same area has deep saline aquifers (>2 km) where they can inject the CO2. At these depths, the CO2 shows little tendency to
come out of solution, and eventually forms carbonates with the local rock matrix.
The net result is 85% less CO2 per pound of coal burned. Due to inefficiencies of the process, it's about 75% less CO2 total. (Takes energy to
compress the CO2.)
Coal gassification is a good step.
Another technology on the horizon is solid carbon fuel cells. They promise an efficiency of 70-80% and deliver a nearly pure stream of CO2 for sequesterization. The higher efficiency and cheaper separation technology combine to cut the CO2 a bunch further.
Clean coal is probably the best answer we have, at least until these ultra capacitors become main stream, and
allow the use of intermittent power sources for continuous loads.
Buy 4 used computers. NOT dells. Not to disparage dells, but they have non-standard form figure power supplies.
Two machiens replace his current machines.
The other two are parts.
Put a second drive in each running machine. Doesn't have to be big, but disks that sit, stick.
Put UPS or at least decent surge suppressors on the working boxes.
I ran a computuer lab in a rural school for 5 years. With 60 machines I lost 3-4 power supplies per year, and 4-5 monitors per year. Lost 5 monitors in 3 days. Surge.
Adding an extra cabinet fan set to vent through 3 removed expansion slot covers is a good idea.
1. Printing.
2. Software. If you are an engineer you likely need to learn Autocad, or a raft of simulation packages. If you are in any data intense environment, you need to use stats packages. If you use math you may want mathematica, maple. If you are an arts major you may need access to Photoshop, adobe illustrator, CAD. If you are a comptuer sci major you may need access to multiple OSs.
Some of this can be handled with site licensing. Much of this can be handled with Virtual machines running on hefty hardware, with the student's laptop being a terminal.
The future of computer labs may be:
2-3 computers for use by students who don't have a laptop, just want to read their email, have to send a file as a genuine Word 2007 document.
The rest of the room are tables with an internet connection. (Bring your own cable.)
At the back there is a printer.
In addition there may be special purpose labs: E.g. E.E. labs where every station has an EEPROM programmer. Media labs where every station has three monitors, and Final Cut Pro.
Compute labs where the machines are simple front ends for working with clusters of computers.
I too could drop my power bill by 30 bucks a year. But I won't do it even for $100 per year.
Reason: State: At present I have 4 browser windows open, each with a dozen tabs, about 40 terminal windows open eitehr ready to run a command or running vi.
I have digikam running on one virtual desktop,
XP running under virtualbox in another desktop.
If I shut down my box I figure it takes me about an hour to get back to where I was.
Given that I do something on my computer 7 days a week, spread out over a 16 hour day, it's not worth it. (I've experimetned with hibernation. Doesn't work well with this hardware. It hibernates beautifully. And never wakes up.
I do let my dual 21" CRT monitors go to sleep.
My wife knows someone at work who comes in, turns her computer on, and goes for a cup of coffee, chats with a co-worker, touches base with her boss, checks her mail box for mail, all waiting while her computer boots and gets all it's ducks in a row. Then she comes in and logs on.
And waits another 3 minutes.
Lets take the 36 buck figure at face value.
Hell lets call it 240 dollars. -- 20 bucks a month.
The average work month has 20 work days in it. A buck
a day. Here the average computer user is likely to get at least $20 per hour. Sure receptionists get 15, but engineers get 40.
A dollar is 3 minutes of employee time.
If we use that 36 buck figure we are at 13 cents per day. Something like 24 seconds.
So tell me again how this saves money?
(Now if you got users to set their bios to start their computer 10 minutes before they arrived....)
Coal plants alter their power levels all the time.
They just can't do it fast. Typically they make about 2% more steam than they need. This handles the transient spikes. As demand picks up they increase the feed rate for the powdered coal, this boils water faster, requiring more feed water. Steam goes to turbine at faster rate, putting more torque on generator shaft to compensate for the increased emf back force from the increased demand.
when the plant is not used at capacity, the parasitic losses (power to keep the plant itself running) are a larger fraction of the total. Less efficient.
When the plant is used at over capacity, you have higher pressure losses in the turbine, hotter steam coming out. Less efficient. But the power curve has a pretty broad top.
Seasonally they put units on and off line.
They can also put a unit in idle mode. It makes steam, it spins the turbine, but there is no load on the turbine, so it takes very little steam. But everything is hot. I think from an idle to full power is only about 2 hours
We have a 10 MW hydro plant on our grid. It generally only releases water during 2 hours during the day, all clustered in the times that power generation is spiking, and most of that in 20 to 30 second spikes.
Unlike a power plant which takes about 48 hours to go from cold to full production, a dam can do this in seconds. When I did the tour there, I could watch the gate actuators for the squirrel cage turbine. They were constantly moving, several times per minute between about 1/4 power and full power. By using the dam this way, they were able to reduce that 2% just in case over production of steam to some smaller number.
The VAST majority of Carbon is in carbonate rocks.
While the amount of carbon is fixed, the fraction of it that is CO2 is not fixed. The fraction of CO2 that is in the atmosphere is not fixed.
There are various cycles that work on different time
scales.
In a temperate deciduous forest there is a
day cycle:
leaf absorbes CO2 during the day converting it
into sugar. Releases CO2 during the night as chemical
reactions continue. The surplus is turned to leaf and
exported to the rest of the tree.
year cycle:
Tree converts CO2 + water into leaf. Leaf falls off tree
and rots, becomes worm food, fungi food.
century cycle:
tree grows up puts on mass. Dies, falls.
ocean cycle:
CO2 is stored in surface water surface water is mixed deeper by wind and current action. In this way the top 400 meters of ocean is at equibrium with the atmosphere with about a 60 year time constant.
ocean deep cycle:
Water mixing to deeper levels seems to take longer. I think I read an estimate of around 12,000 years.
rock cycle:
Carbonate rocks subducted at edge of plates. Magma heat
diassociates carbonates. CO2 released in volcanoes.
CO2 taken up by algae, fall to bottom of sea. Time and presure form carbonate rocks. Residence time millions of years.
So long term solutions lie in understanding the residence time of these various cycles, and giving them a nudge toward the pools that make it a less immediate problem.
Conifers are darker than grass. Plant large areas to conifers, and you lower the albedo of the planet. Which
increases the temperature.
Increasing the temperature at arctic latitudes, allows the tree line to move north, taking over relatively light areas of gravel and peat bog and turning them into dark forests.
This increases global warming...
Putting wood and paper in land fills actaully does do a pretty good job of sequestering it. Most landfills are anerobic, and while there is some methane production, most of the carbon just sits there. Saw a story about 50 year old newspapers pulled from an LA landfill still quite readable.
An even better way to sequester carbon is to turn it into charcoal, and mix it into the soil. Charcoal has a very long residence time in the soil, and acts as a sponge for nutrients, reducing leaching.
I have a Pogo Linux workstation. (Quiet!) with a dual core chip and 8 GB ram.
I run VirtualBox with a copy of WinXP for those applications that there isn't a good linux implementation
for. (Sorry -- Gimp != Photoshop Glom != access) It's allocated 1.5 GB ram, and mounts my home directory from the real machine via Samba. I get acceptable performance on everything I do on it.
For test bed type applications, I think that RAM is your biggest consideration. You don't want the VM to swap to
virtual memory. (VMVM? VM2?)
My brother tells a story of a bio grad student who was hired to teach biology for the last half year when the original one quit.
Because he didn't have a teaching degree, they moved all the 'real' students to to the other teacher, and he got all the students who 'needed a science to graduate before they go work in the mill'.
Most of these kids took shop too.
So he walks into his first class after spending all night going through dumpsters in town, gives each kid a large tin can and a mouse trap, with the assignment, "Build a live trap."
They spend the rest of the period brainstorming approaches.
Within a few days he has many live traps.
"What do mice like to eat"
Many answers.
"How could we find out"
Much discussion, but it boils down to, try an equal number with various food and see what you get.
So they do that.
Then he starts looking at where the traps were. And this leads into doubt about the first experiment. Some guys put their traps in the open, others by the wall. Some on the school property, some in their back yard.
More discussion.
More experiments.
Then he asks for what results they got last week. Nobody is writting stuff down. This turns into the importance of record keeping.
Then he asks, " Are you catching the same mouse over and over, or are you catching different mice?"
Whole can of worms about identification. They figure out a way of marking small diameter aluminum tubing and gently crimping it as a bracelet. Now they can tell if a mouse has been caught before.
"How big an area does a mouse forage?"
So they took over wild section of land behind the school, and placed the traps in a fairly even distribution over several acres. Most of the kids had between 6 and 12 traps now, and were visiting them twice a day.
Each kid recorded every mouse they caught, adding a new band when an untagged mouse fell into their clutches.
By this time the weather started getting nasty, as fall gave way to winter.
So the last few weeks to Christmas were spent analyizing the data. So he taught them about averages and means, and standard deviations, and regression testing. (This before calculators. All hand work.)
Remember this is a class that 'can't learn much, we can only hope they will be good with their hands'
Just before Christmas, the school imspector came in to observe. He looked at the neat notebooks, saw the earnest work the kids were doing, had to ask the teacher what a kid was talking about when mentioning mortality tables, and mean forage diameter differential between male and female mice.
The grad student explained.
The school inspector said that he wasn't following the curriculum, spoke to the principal, and had the grad student fired.
Open DNS is wonderful. I used it at the school where I worked. It didn't stop the porn, but it cut the incidence of accidental porn to nothing.
Also: Set google to "moderately safe"
The other point about having the computer in a public space is good. Doesn't have to be in the living room next to the noisy TV, but in a corner where people walk by on a frequent basis is good. And don't then use YOUR computer in a private space. Set an example.
I also think there is merit in doing homework on the dining room table. Folks are around to help. 'Room time' happens after 'chores' such as washing the dishes, and homework are done.
Still, they are going to run into content that is disturbing -- either new ideas, (Muslim family law -- role of women, sexual freedom) violence, bigotry, hate, and in general man's inhumanity to man. Start early and talk about what she's seen. Get her in the habit of asking questions to you. These discussions are springboards for talking about ethics and morality in general.
You have to decide about social networking sites. Much of that use depends on the crowd she falls into.
Remember too, that even if you control her access she will have less restricted access at friends' houses, at school, or even the public library. Teach them responsible use. Teach them to be skeptical, to figure out ways to verify, to look for consistency. Done this way you have a huge impact on her development as a responsible adult.
Somewhere around 10 MP you get to the limits of resolution for the typical reduced frame (2/3 the size of a 35mm frme) digital camera.
By extrapolation 16 MP would be around the limit for a full frame camera. (And then you have to buy a new set of lenses)
If you aren't pro you likely don't maintain your lenses clean enough to reach these resolutions. Indeed few lenses reach anything like theoretical resolution (Raleigh criteria)
I've got a Nikon DX70 which I love, but ...
* It's performance in low light is noisy.
* It only accomodates about 6, maybe 7 stops of illumination variation in a scene. Reminds me of shooting ektochrome slide film. Expose for the highlights, or for the shadows. You can't have both. Unless you use flash to fill in. By comparison film accomodates about 9 stops, and can be tweaked to 10-12 in the darkroom. (Much of the Zone system of the black and white photographers is based on these tweaks.) 10 stops is typical for a scene on a cloudless day.
The future:
With fast enough data collection you dump the CCD at 1000 frames per second, then post process for both noise reduction and image stabilization. Indeed: Collect during the time that the diaphram is closing and you may be able to get advantages of the speed of a wide open lens and the depth of field of a stopped down lens.
I suspect to do this you would have to build the processing into the chip. One proccessor per NxN pixel square? Each processor hands data back and forth to it's neighbors for handling image stabilization?
Set your camera up on a stand looking at a grey card. The camera takes a series of frames to calibrate the individual sensitivity of each sensor.
Take off list. Email me at sgbotsford@gmail.com
Framemaker.
I have a version of the 5.56 beta released for Linux that I hacked to remove the time limitation.
It is *still* the best way to write any document that's more than 5 pages long and using more than 4 styles.
The equation edition is not as fast as latex, but much faster than word's equation editor. Results are between the quality of word and the quality of latex too.
As to *TeX. All is well if you want a standard document on a standard page. Just try to be creative with layout. E.g. A 20 page document that has 18 figures, and your style guide is that figures are either at the top or bottom of the page, but if there are two figures on facing pages, preference is that one is on top of the column, and one on the bottom, and that if the figure is on the top of a column, there should be a .5 pt line, 75% of the column width between the caption and the main text.
In a former life as a teacher, I went to the headmaster and said that I wanted to teach the grade 10 general science course differently.
I wanted to walk in the first day and talk about 4 elements, earth, air, fire and water. Then for a day show how this explains everything.
I wanted to teach about caloric and phlogiston and luminiferous ether, and with each show the explaining power they had, then the problems they ran into.
Ptolomaic astronomy with circles and epicyccles and eccentrics.
Geology would be fun, especially the bits with continental drift and catastrophism vs uniformism.
The goal, I explained to my head: Teach that science is a process, and not a body of knowledge.
Long silence.
"That is a very dangerous idea."
I don't work there anymore.
With virtual box you can make an image 'immutable.' Rebooting the virtual machine reverts to last state.
Use samba to mount a folder from the host machine to store your class stuff.
You can also set up the networking so that the virtual machine has no network contact with any computer but the host, and the world.
Couple this with a reasonable firewall, and you're set.
Hmm. Let's see. Consider a 1 meter long rail gun.
v is final velocity
a is acceleration
m is mass of the projectile.
W is power.
v = at
E = mv2/2
But
E = Wt
Equating these two Wt = mv2/2
v2 = 2Wt/m
But t=L/2v (from average velocity v(ave) = v/2)
v2 = 2WL/2vm
v3 = WL/m
v = cube root (WL/m)
So a 8 ram bullet ...
v = cube root (25000/.008)
= 146 m/s or about 500 ft/sec
Bit slow yet.
To get up to 2000 feet per second will take
64 times the power.
Won't worry about it this week.
The batteries may make recharging the capacitors
faster.
Let's see: to get a 700 m/s bullet (2200 fps)
E = mv2/2 .008 * 700^2 / 2
E =
= 2000 joules.
Your 1 liter battery could provide power for 12 bullets
per second. Not too shabby.
Much of this depends on what you want to do.
If you want to run the Big 4 (word processing, spreadsheet, presentation, internet browsing) then FOSS solutions are idea. In my last school however:
* Staff used a client server commercial package called RenWeb for record keeping, lesson plans, and parent communication. Client was winsooze only .NET application with a raft of custom DLLs
* As the photography instructor I used and taught photoshop. I'm sorry but the Gimp just doesn't cut it.
* As an outdoor program coordinator I used Mapmaker Pro and Oziexplorer to build maps and to move data to/from my GPS. I've found no equivalent of Mapmaker that is OSS.
* As part of support we had a number of MS access quick and dirty database apps, each with a bunch of entry forms, and reports. I've spent days searching for an equivalent package that allows rapid database application prototyping. (I looked at kexi, knoda, rekall, bond, glom, pfm, and PgAdmin)
This doesn't mean that FOSS is not usable, but in most systems it will need to be a parallel system to paid software.
IF I were in charge of a school district I would look at doing it this way:
1. I would deploy student machines as being some form of terminal only. Students with laptops could use windows rdesktop, or VNC.
2. I would buy high end boxes stuffed with memory and run a bunch of instances of virtualbox or equivalent vmware, or remote X. Using immutable images in virtualbox makes for systems that are hard for the kids to corrupt.
3. Users on startup could select if they needed a windows session or a linux session or a mixed session. Tweak the system so that equivalent applications on linux ran faster.
4. Servers would run FOSS. Working around Active Domain isn't that difficult, especially if the vast majority of your windows installations are virtual.
5. In a few cases you need individual high powered machines. (Running photoshop in virtualbox is less than satisfactory)
I would make the transition as follows:
A. Using whatever machines I could get, I would start setting up ancillary servers. DHCP, YP, SMTP, POP, IMAP, NFS, SMB. This gives my admins experience in working with open source operating systems, but in a way that is easy to back out of. E.g. Initial Samba services can be for doing backups.
B. When this works, and admins are comfortable with it, I'd bring in 1 high power box per site, and set it up as a VB server. Initially it would serve only windows, and it would serve to visiting laptops. This is sold as a security measure to protect your network from laptops with unknown software.
C. Once this works, I'd convert one lab at each site to thin client setup. Their existing hard drives would be untouched while learning how the system works. Eventually this is sold as a cost cutting measure, as it permits running the lab machines for more years. After the admins are happy with the results, rip out the disks. This makes the lab both quieter and cooler.
D. Introduce VBox with saved state to staff. Being able to shut down in their classroom, and reopen at home and have the same machine state for their record keeping, lesson plans and so on will be a win. Or they can leave the machine at work, and have a client session from home.
E. With the money saved from not having to upgrade all the labs, I'd make a few high end machine sets for applications that are demanding such as Photoshop, and Final Cut Pro.
Depends on your environment. When I worked for a university department, I had a chronic problem with grad students hooking up all sorts of things to the ethernet ports. And with cheap ethernet devices this is a security concern.
We finally got switches that could be told, "This port belongs to this MAC" and outlets that weren't in use were not connected between the patch panel and the switch.
All this meant was that the grads had to learn how to clone the MAC address of the device they were replacing. Then nmap would tell me that the Linux box had turned into a Windows box.
Nowadays I expect that they would put in their own $40 wireless AP. Once a dozen of them did this, no one would be able to do anything on the air.matchedmatchedmatchedmatchedmatchedmatchedmatchedmatched
At one point I was in the middle of a transition between 10Base2 and 100BaseT networking. All done on a shoestring budget, with no permissible downtime. My wife bought me a Dilbert T-shirt. Dogbert is walking out of the server room, the room a shambles. He's carrying a baseball bat. "The Network is down, but I'm feeling much better."
I wore that shirt to work several times over the next few weeks. People gave me a bit more space and less hassle about the constant, "Excuse me while I run 11 cables above your office ceiling" interuptions.