Quanta Building MIT's $100 Laptops (eWeek) [LWN.net]

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 0:50 UTC (Fri) by rknop (guest, #66)
Parent article: Quanta Building MIT's $100 Laptops (eWeek)

The networking cards in these things-- are they gonna have proprietary drivers, or are these laptops something that (in principle) one could install Debian on without having to download nonfree stuff?

I haven't read enough to know if they'll have USB or any way of getting a CD on there-- I'm guessing not. But if they *do* have a CD drive, then, heck, these are useful general purpose computers. Slower than most today, but at $100 a pop, I'd be tempted to pick some up for my work, even though I'm not in a poor and disadvantaged country.

-Rob

(Log in to post comments)

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 1:15 UTC (Fri) by bk (guest, #25617) [Link]

According to http://laptop.media.mit.edu/ , the machines will have primarily black and white displays (I'm assuming they mean grayscale) and are flash based (no hard drive).

I agree that these can be attractive general purpose machines provided they: a) Are significtanly smaller and/or lighter than full featured laptops, and b) get significantly better battery life. I'd like to see a machine that has the battery life of my mp3 player, something like 10-15 hours.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 1:20 UTC (Fri) by rknop (guest, #66) [Link]

I'd be surprised if they were significantly lighter-- you usually pay for that.

From the article, I'm not sure if they really mean B&W displays-- it may be that it's only B&W under bright sunlight, but it sounds like there may be a color mode when lighting is more reasonable?? The advertising pictures show color screens.

Also looking at the article, it looks like they won't have standard networking, which would rule them out as (even) dumb terminal-type workstations for my kind of use. (Of course, for that, I can in principle use recycled 500MHz desktops, if I can find them.)

-Rob

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 4:34 UTC (Fri) by dlang (subscriber, #313) [Link]

no, they're planning electronic paper displays, which are B&W (except for the bleeding edge versions), in fact the common ones I would expect them to use are only 4 bits or so of greyscale.

given the planned use, I would be surprised to see a cat-5 network interface (it's a way to get gunk into the machine and wires to break), but I would expect some sort of slow-but-reliable wireless capability (not nessasarily 802.11, bluetooth would work for it)

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 8:00 UTC (Fri) by irios (guest, #19838) [Link]

They will all have WiFi, in some sort of self-weaving ad-hoc mode that does not require access points or manual configuration.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 11:55 UTC (Fri) by dlang (subscriber, #313) [Link]

they will have some form of wireless networking as you describe, but will it be 802.11 based (as implied by the term WiFi) or something else?

going with 802.11 has the advantage of interoperability with othe systems (but how you get the self-weaving ad-hoc mode without conflicting with other uses of 802.11 and being able to interoperate with them is tricky)

going with something else could still result in interoperability if there are access points available (just like you can use bluetooth to connect to a cell phone, you could have a cheap box that is a gateway to a landline, or satellite phone. you only need one per community for everyone to share)

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 12:30 UTC (Fri) by Duncan (guest, #6647) [Link]

The sources I've read (several of them in the couple months of coverage
I've seen) indicate a dual-mode display, similar to what's in many
portable DVDs today, color at something like 640x480 under indoor
conditions, black&white at 3X the resolution where needed, as in bright
sunlight.

The networking will be ad-hoc wireless mesh network built-in, no Ethernet,
altho they have several USB ports so you could use USB/Ethernet adapters,
if desired. (Note of course that these adapters would add significantly
to the cost, at the prices we are talking.)

One source I read, possibly unreliable, stated that Apple was willing to
donate a version of their operating stack, but it was rejected due to the
strings attached -- lack of truly free licensing. If that's the case,
then the wireless networking will likely be decently free. With the
volume they are looking at (double- to triple-digit millions, block
purchases of a minimum of a million at a time, minimum 10 million ordered
to start production), they could hard-wire the connection frequencies such
that the laws blocking full source availability in some countries wouldn't
apply, since the hardware couldn't be programmed outside its allotted
frequencies in any case.

Power will be flexible, including a hand-crank good for 20 minutes
operation, and batteries, probably C or D cell, widely available
worldwide. Other specs include 128 MB DRAM memory, 1/2 a gig flash memory
non-volatile storage (no hard drive due to expense and fragility), several
USB ports.

Note that several stories have mentioned non-conventional uses as well.
Preliminary versions used in parts of SE Asia have been in places without
electricity, so the laptop ends up being the brightest light source as
well as the only computer in the home. Consider it operating in B&W mode
with an all-white screen, hand-cranked a couple minutes every 20 to
provide light for the home.

The computers should also be available retail, altho at significantly
higher prices, likely $200, possibly $300. One can imagine Walmart
ordering a million of these and selling them at $200-250 each, $250-300
including an accessory kit with an AC adapter, a USB/Ethernet adapter,
perhaps a USB patchcord and MSWormOS computer sync software, etc. (One
imagines these operate as USB-hosts, so that patchcord would need to be a
host to host patchcord, probably with a box in the center with two
USB-gadget interfaces, thus, not just a simple USB cord. Unless the
computer comes with say four USB host ports and one USB-gadget port, like
those old Ethernet hubs with an uplink port.)

Duncan

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 19, 2005 9:53 UTC (Mon) by ekj (subscriber, #1524) [Link]

Did you ever wonder how come, when they invented usb they repeated that braindead error of ethernet where there's two different sorts of cable, one crossed, and one non-crossed where it's hard for an untrained person to see, or understand the difference ?

With usb it's not the cables, but the simple fact that you cannot simply connect any two devices equipped with usb -- one requires one "host" and one "gadget", rather than the usb-bus being completely neutral.

USB

Posted Dec 20, 2005 8:58 UTC (Tue) by Duncan (guest, #6647) [Link]

USB was defined by Intel, back when x86 CPUs were first going over about
500 MHz and they were searching for ways to use more CPU, so they could
continue to sell faster chips, given that 500 MHz was really the level at
which the CPU was "fast enough" (on x86) for what most folks did with it,
for the first time in history.

Thus, USB is a CPU intensive hosted protocol, requiring a CPU at the
"host" end, altho it requires far less at the "gadget" end. Compare it to
the "Fire Wire" of the time, a similar usage protocol, similar cables, but
MUCH faster, and using a peer2peer protocol rather than the hosted
protocol of USB. The p2p interface required a more expensive interface
chip in each item, but as it was p2p rather than hosted, it didn't use as
much CPU and all ends were the same. The additional incremental expense
for each gadget, in addition to the fact that Apple tried to keep the Fire
Wire name proprietary at first, and the difference between Apple and
Intel, was the reason USB became popular much faster than Fire Wire.
(Apple eventually relented and yielded the Fire Wire name to the industry
group defining and controlling the standard, but it was really too late by
then, and the incremental cost add of the interface for gadget makers did
nothing to help.)

Duncan

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 19, 2005 9:48 UTC (Mon) by ekj (subscriber, #1524) [Link]

They'll have 802.11 of some flavour or other, combined with some sort of software to create ad-hoc mesh-networks when in the vicinity of eachother.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 1:16 UTC (Fri) by khim (subscriber, #9252) [Link]

The networking cards in these things-- are they gonna have proprietary drivers, or are these laptops something that (in principle) one could install Debian on without having to download nonfree stuff?

I'm pretty sure it's not yet undecided. But since Linux is used mainly to reduce costs... I do not think abyone will think twice about using firmware loaded from main storage (it'll reduce costs) - and that alone will make Debian uninstallable, right ?

haven't read enough to know if they'll have USB or any way of getting a CD on there-- I'm guessing not.

Of course there will be no CD: they are using flash instead of HDD to improve surviveability in harsh environment and if you'll compare poternitial problems of HDD and CD... CDs were pretty much ruled out from the very start. USB is planned, of course...

Why the hell noone point to developer's web-site ?

Slower than most today, but at $100 a pop, I'd be tempted to pick some up for my work, even though I'm not in a poor and disadvantaged country.

$100 is for batch order of minimum 1'000'000 notebooks. AFAIK the plan is to sell for $200 if you want just one or two. Still interesting, but not as notebooks: they can do most things my Zaurus can do BUT they must be 10 times more reliable. And they can be used where electricity is a problem! If they'll keep $200 tag for retal version I will probably buy one: this thing is significally bigger then Zaurus yet small enough for me and way more capable then Zaurus... and few times cheaper even with $200 pricetag...

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 1:22 UTC (Fri) by rknop (guest, #66) [Link]

Only if the firmware is closed source. If there is actual open source firmware somewhere, Debian would have no trouble incorporating that into the distribution.

-Rob

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 1:28 UTC (Fri) by khim (subscriber, #9252) [Link]

You can be 200% sure it'll be closed source. Regulations in some countries require it.

P.S. Actually they require "hack-proof" design (to stop misuse of radio frequencies) and binary blob is not exactly hack-proof - but it's hack-proof enough to satisfy bureaucrats. Open source firmware is not hack-proof enough even for them...

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 7:55 UTC (Fri) by irios (guest, #19838) [Link]

What?

There ARE widely available wifi chipsets out there with free drivers (at my home, Intel, Ralink)

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 11:09 UTC (Fri) by khim (subscriber, #9252) [Link]

Do not mix free driver and free firmware. Yes, situation with free drivers is better today, but situation with firmware is not: most free drivers for cheap cards include some "blob" to be loaded in card on startup; more expensive models include flash where firmware is stored - then driver is truly free.

Since blob is in fact computer program (even if for specialized CPU and compiled with specialized, usually not freely-available, tools) Debian claims it's not enough to have this blob and open-source driver: you need sources for firmware too. Not something ODM will ever bother with: it's not even executed on CPU, it's not related to Linux is any way (the only relation is that Linux system must have this blob to initialize card), there are no need to ever port it (even if you'll put card in your Mac with PowerPC main CPU specialized CPU on card itself will be the same), so noone except Debian are bothered by that...

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 14:00 UTC (Fri) by vonbrand (subscriber, #4458) [Link]

Fedora doesn't ship the firmware for the intel ipw2100 and 2200 WiFi cards either, for exactly the same reasons as Debian.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 21:14 UTC (Fri) by drag (subscriber, #31333) [Link]

No!

Fedora and Debian can't include the Intel Firmware because Intel forbids redistribution of it. Debian maybe stinkers about it since most firmware forbids modifications and such, but they'd still distribute it in non-free.

It's Intel that is the problem, not Debian, not Fedora.

Also cards that don't require firmware are not nessicarially more expensive. Ralink rt2500 style cards do not require a loadable firmware to operate and they are _still_ cheaper AND they have fully open sourced drivers provided by the manufactures!

Go ask the OpenBSD people if you don't beleive me.

Any distro offering firmware images as part of their normal install for Intel wireless cards can face a lawsuit from Intel unless they have a prior licensing agreement, which you cannot use youself.

Also:

Firmware != Binary blob.

Firmware is not the same as compiled code. It's a specific item for a specific peice of hardware and is completely architecture and OS independant.

You can use the same firmware images for Linux on PowerPC, on Linux on x86, on OS X, on Windows XP, on OpenBSD. It doesn't matter.

Binary blobs are kernel code, based on Linux kernel developer's code, compiled from C source code before they are made aviable to you and when they are used become part of the Kernel code running on your system. You can't use the same binary blob on a x86 Linux as on a PowerPC linux, for instance.

There is some 'glue' code that people provide with binary blobs, but it exists only to shoehorn generic kernel code into a specific kernel (aka the one your running).

There is a huge difference between binary blobs and actual firmware.

Beware that some manufacturers want to cloud the issues or mislead people by giving the name 'firmware' to things that are not firmware. For instance on the Linksys they call the OS that they load on the onboard flash drive 'firmware upgrades' when in reality it's just a striped down Linux or Vxware OS.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 18, 2005 4:42 UTC (Sun) by khim (subscriber, #9252) [Link]

For instance on the Linksys they call the OS that they load on the onboard flash drive 'firmware upgrades' when in reality it's just a striped down Linux or Vxware OS.

If it does not run on main CPU - it's still firmware. It does not matter if it's full-blown OS or not. If it's Linux - you are entitled to sources, if it's Vxware OS - you are not, but either way from PC's point of view it's still firmware.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 11:10 UTC (Fri) by tajyrink (subscriber, #2750) [Link]

Free drivers, maybe, but not free firmware if they require such to be uploaded from the host computer. And eg. Intel doesn't have the firmware even freely redistributable, like Zydas, Atmel and some others do (not that it'd help including it in Debian main, but at least they could be provided from more places).

Educated guess

Posted Dec 16, 2005 12:18 UTC (Fri) by man_ls (subscriber, #15091) [Link]

I think that the legal issues facing wi-fi are as follows: 802.11g establishes a range of frequencies and mechanisms for hopping between them, in order to avoid any resonances or dampings and find working frequencies in any environment. The range of frequencies varies from country to country: US, Japan, maybe even within Europe.

The cheap solution is to have one chip and change the range of frequencies in software (in this case, firmware). If you have access to their sources, you can reprogram the chip to go outside the range of legal frequencies. Probably not a big deal; but companies may face liabilities.

A good, expensive solution would be to have different versions of the chip for the different markets. Another one might be to limit yourself to a range of frequencies which is common to all countries (the intersection between all ranges); I don't know if this is at all possible. So probably Intel and Ralink are doing one of these things, and their chips can be safely controlled in free software.

This is all an educated guess. Can anyone add a more knowledgeable opinion?

Educated guess

Posted Dec 16, 2005 14:20 UTC (Fri) by khim (subscriber, #9252) [Link]

Essintionally correct. Just one change: noone ever goes via "different hardware for different countries" route- unless difference in frequency is so big they need to redesign high-frequency part (like GSM 1800 vs GSM 1900). But if adapter is not extremally cheap they just put flash on card and store firmware there - so no need to keep firmware in driver.

What I can not uderstood is why this is such a big deal. Non-free firmware is always in your system: in your WiFi card, in your CD-ROM, in your HDD, etc. And it's not a big deal for Debian creators. Why is it such a big deal if it's on CD in and not on chip then ?

Educated guess

Posted Dec 16, 2005 17:41 UTC (Fri) by bk (guest, #25617) [Link]

Distribution, obviously. Debian doesn't distribute the firmware in your HDD or DVD drive, but it does have to distribute non-free binary blobs.

Educated guess

Posted Dec 17, 2005 12:51 UTC (Sat) by man_ls (subscriber, #15091) [Link]

Yeah, it is copyrighted material: we would be at the mercy of the vendor, who might limit distribution if it wanted e.g. to sell new wireless adapters at any point in the future.

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 1:20 UTC (Fri) by nlucas (subscriber, #33793) [Link]

http://laptop.media.mit.edu/faq.html

Quanta Building MIT's $100 Laptops (eWeek)

Posted Dec 16, 2005 15:42 UTC (Fri) by danieldk (subscriber, #27876) [Link]

"built in 802.11 wireless (probably a Athleros chip, due to its advanced driver), with mesh networking."

Source: http://www.gettysfamily.org/wordpress/