Today, our computers are always on, never restarted,
always connected.
This is a bad thing, not a good thing.
My personal machines and devices are not high-availability servers, and don't need to be powered on all the time. I close my browser windows, and power off my machines frequently, and deliberately. I generally don't gain very much from the practice of leaving the room for an hour, with my machine powered on.
I'd like to dissent from this echo chamber of constantly advocating promiscuous systems, that tend to do things for us without asking us first, and promote biased decisions in favor of permissiveness. I pretty much hate that you can't remove the battery from an iPhone.
That said, there was an time when client/server computing delivered a lot of the kind of functionality, via a fragmented industry of proprietary "enterprise" applications, that has, lately, been re-cast as the broad ecosystem of "mobile apps" (which are essentially the same thing, only smaller and portable). Also, since HTTP is a stateless protocol, "rich web applications" (aka: HTML5+ & JS), have been able to fill in a lot of gaps in web browser behavior with asynchronous requests. I tend to think that JavaScript will become the open version of the once proprietary non-interoperable model of native binaries and client/server computing.
For you, perhaps. I use my open browser tabs as a current "state" of my mind. My todo list, if you will. You might use a different method for keeping track of what you have to do. Maybe you have a good memory.
At least the browser tabs come back after the browser quits. My main reason I never reboot is my terminal tabs. Each one is cd-ed to a project directory and each has command line history and scrollback history that are potentially important reminders of where I am on that particular project. I go for weeks without rebooting (even in the face of OS updates) because I don't want to lose my terminal tabs.
Powering off machines is a detail. They should go into low power mode and switch themselves off when they are not being used. Why should I have to remember to do it? They are machines, they're built to remember things like that. As long as they save their state, it doesn't matter.
Yeah, a lot of this is about personal preferences, and preserving them across time and space. It's not always a bad thing. There's always a trade-off between convenience, automation, predictability and security.
Another good point spolu brings up:
The chrome/ subdirectory of the chromium project (all
that makes up the Google Chrome experience: tabs, settings,
sync, omnibox) is made up of 5,343 C++ implementation files,
adding up to 1,449,451 lines of C++ code (as of 2013/09/11).
It's hardly tractable by one person, and hints to the fact
that it's probably impossible for a one guy to modify it
alone to come up with a profoundly different experience.
This detail doesn't get a lot of discussion, but I think it's sometimes an essential fallacy of the very premise of the open source concept, that having the ability to audit source code, doesn't necessarily make it a realistic option.
Lots of modern contraptions are simply too complex for any individual to ever claim a hope of gaining an honest understanding of their inner workings, nevermind an unskilled layman. At least, not without living like a cleric, cloistered away in the library of some convent or monastery for most of one's life. Sometimes I feel that the real hazard is that we gloss over the idea that we're trying to exert control over a massive arrangement of billions of transistors, however small they may be. If we were confronted with an elaborate array of ordinary light switches, and told to conquer the same tasks, I think it'd be easier to realize the true scale of some of the things we fiddle with on an everyday basis.
It's more than just the ability to audit, it's the ability to fix specific problems so you can focus on your business requirements. This happened to me recently -- I was using Apache in a rather unusual reverse proxy configuration and was unable to suppress (using the usual directives) certain headers which were causing problems upstream. I could have spent a long time researching other solutions or coming up with clean workarounds but at the end of the day it was quick and effective to simply edit mod_proxy_http.c and fix the problem directly. Now that the service is up again, I have time to come up with better solutions if I desire.
When our closed systems misbehave, my only option is to shrug and ask if we have a support contract.
> an essential fallacy of the very premise of the open source concept, that having the ability to audit source code, doesn't necessarily make it a realistic option
I don't think the studious hermit going over every line of open source code was how it was supposed to work either though.
It's more just that you have lots of groups exercising the code and looking into various portions of the code - if every group decides "Yep, the portion we use/looked at is solid", then you gain some confidence that the code does what it's supposed to. The more groups using/perusing the code, the more confidence you gain.
there really is no trade off between "convenience, automation, predictability and security."
there's nothing magic about switching your machine off to increase security. It's just as vulnerable when you're using it, more so in fact. I'm not sure how shutting your machine off adds any predictability either. If you want to achieve the same thing just turn off wifi.
For mac laptop users we get the best of both worlds, always state preserving, but essentially "off" when not in use. Sadly neither windows nor linux have ever got "sleep" right.
I do the same thing with terminal tabs, is there a program out there that remembers the working directory and history of the terminal tabs that are open, then after closing reinstates it all when opened? This doesn't seem like it would be too hard to do.
Kind of like a tabbed browser that remembers history and open pages, but for terminals.
(1) screen will do a lot of that. But if the system has been restarted, live processes in the terminals will no longer exist.
(2) You have the option of a suspend-to-disk, which can 'wake' you back to the right place.
(3) You could use an OS running in a VM, and suspend that. This is arguably the cleanest solution right now.
(4) You could get really in to process checkpoint/restore stuff in the kernel which seems to be constantly under development to facilitate live process migrations across machines in data centers.
Actually though, in all cases you will lose your network connections, which might be critical to the processes as well. Perhaps re-evaluate your workflow.
I use iTerm on Mac and it has a setting you can enable that remembers tabs and even opens new tabs to the current directory of your current tab. When I restart my computer, the tabs and command history for each are preserved, but the output display is lost.
Well, it's not powered all the time, your computer probably go to sleep while you're not in the room for a period of time, but the state of the computer is kept in memory. So it definitely feels like it never stopped. Having that state kept around is hardly a bad thing IMO.
I believe that classification would be more suited for a *BSD user or someone concerned with integrating the Plan 9 userspace into their workstation.
Linux has a pretty low entry point nowadays. Plenty of people I know surf the web, use Skype and check their email like any other user but do it on Ubuntu.
There are some WMs which either emulate look-and-feel of the Windows environment, or let you map hotkeys, that this isn't too much of a hurdle. My preference happens to be WindowMaker, which allows extensive hotkey mappings.
Have I done something weird to my Mac? 10.7.5 Lion here, when I press the power button it comes up unchecked by default.
"Edit: since 10.7.4, the "Reopen windows when logging back in" checkbox has stayed unchecked if you uncheck it once, so the hacks below are not needed anymore."
I think there are some stateless SSH / terminal options you can use.
screen, certainly (I rarely ssh to a host without starting screen), though tmux is what the cool kids are using these days.
I regularly suspend my laptop. Almost as good as leaving it powered. The main hassle is if you've got multiple SSH sessions to the same host, and you need to sort out which is what.
> I'd like to dissent from this echo chamber of constantly advocating promiscuous systems, that tend to do things for us without asking us first, and promote biased decisions in favor of permissiveness.
This is a decent argument (partially) in favor of the ideas discussed in the OP. I'd love to see a world where browsers are akin to libraries, and hackers roll their own.
> I pretty much hate that you can't remove the battery from an iPhone.
This is why I don't own one. I encourage you to vote with your wallet.
You explain why it's not necessary for you, but you don't explain why you think it's a bad thing. Why is it bad that our computers are always on, never restarted, always connected?
As long as your monitors power off and you're not leaving the system doing much CPU work, modern computers don't use a lot of power when quiescent. Sure, they use more than being turned off, but they're not exactly IR hotspots.
For comparison, as much as your washing machine, dishwasher and all your 10 (cfl) ceiling lights together (from http://www.carbonfootprint.com/energyconsumption.html). Or as much as a low efficiency fridge-freezer.
Multiply by 2 if you run air conditioning. Multiply by number of employees for an office. Etc.
Your numbers seem off. 7-year-old desktops with monitors off rate 50-70 watts. More recent desktops in that link are 30-40 watts, so you're looking at numbers a little bit more than 20% smaller.
I'm also not sure about your office, but in my office the lights are on for much more than the four hours per day in your other link.
And similarly, we're talking about when I'm not using my computer for work, so it's not 440 kWh/year (24x50x365). Given I work about 9 hours a day, the portion of a year that I'm not at my workstation (including weekends) is about 250 full days - so the number is actually more like 240 kWh/year (24x40x250) for leaving it on overnight.
My personal machines and devices are not high-availability servers, and don't need to be powered on all the time. I close my browser windows, and power off my machines frequently, and deliberately. I generally don't gain very much from the practice of leaving the room for an hour, with my machine powered on.
I'd like to dissent from this echo chamber of constantly advocating promiscuous systems, that tend to do things for us without asking us first, and promote biased decisions in favor of permissiveness. I pretty much hate that you can't remove the battery from an iPhone.
That said, there was an time when client/server computing delivered a lot of the kind of functionality, via a fragmented industry of proprietary "enterprise" applications, that has, lately, been re-cast as the broad ecosystem of "mobile apps" (which are essentially the same thing, only smaller and portable). Also, since HTTP is a stateless protocol, "rich web applications" (aka: HTML5+ & JS), have been able to fill in a lot of gaps in web browser behavior with asynchronous requests. I tend to think that JavaScript will become the open version of the once proprietary non-interoperable model of native binaries and client/server computing.