www/about.html - third_party/toybox - Git at Google

 <html><head><title>What is toybox?</title>
 <!--#include file="header.html" -->

 <h2><a name="what" />What is ToyBox?</h2>

 <p>Toybox combines the most common Linux command line utilities together into
 a single <a href=license.html>BSD-licensed</a> executable. It's simple, small, fast, and reasonably
 standards-compliant (<a href=http://opengroup.org/onlinepubs/9699919799>POSIX-2008</a> and <a href=http://refspecs.linuxfoundation.org/LSB_4.1.0>LSB 4.1</a>).</p>

 <p>Toybox's 1.0 release goal is to turn generic Android into a
 development environment capable of compiling <a href=http://www.linuxfromscratch.org>Linux From Scratch</a>.
 A tiny system <a href=/aboriginal>built from</a> just toybox, linux, <a href=http://musl-libc.org>a C library</a>, and a C compiler (such as LLVM or
 gcc 4.2.1+binutils 2.17) should be
 able to rebuild itself from source code without needing any other packages.</p>

 <b><h2><a name="status" />What commands are implemented in Toybox?</h2></b>

 <p>The current list of commands implemented by toybox is on the
 <a href=status.html>status page</a>, which is updated each release.
 There is also <a href=roadmap.html>roadmap</a> listing all planned commands for the
 1.0 release.</p>

 <p>In general, configuring toybox for "defconfig" enables all the commands
 compete enough to be useful.  Configuring "allyesconfig" enables partially
 implemented commands as well, along with debugging features.</p>

 <p>Several toybox commands can do things other vesions can't.  For example
 the toybox "df" isn't confused by initramfs the way other df implementations
 are. (If initramfs is visible, df shows it like any other mount point.)</p>

 <b><h3>Command Shell</h3></b>
 <p>The Toybox Shell (toysh) aims to be a reasonable bash replacement.  It
 implements the "sh" and "toysh" commands, plus the built-in commands "cd" and
 "exit".  This is the largest single sub-project in toybox.</p>

 <p>The following additional commands may be built into the shell (but not as
 separate executables): cd, exit, if, while, for, function, fg, bg, jobs, source,
 <a href="http://opengroup.org/onlinepubs/9699919799/utilities/alias.html">alias</a>,
 export, set, unset, read, trap, and exec.  (Note: not done yet.)</p>

 </ul>

 <h2><a name="commands" />Which commands are planned?</h2>

 <p>The toybox <a href=todo.txt>todo list</a> mentions many potential commands
 which may be added to this project.  (Whether that file is readable by anybody
 but the project's maintainer is open to debate.)  The roadmap wiki in the
 nav bar has a more human readable version.</p>

 <p>The criteria for a toybox 1.0 release is that a system built from just the
 Linux kernel, toybox, C library (such as uClibc), and a compiler (such as
 tinycc) can rebuild itself from source code.</p>

 <b><h3>Relevant Standards</h3></b>

 <p>Most commands are implemented according to
 <a href=http://opengroup.org/onlinepubs/9699919799/idx/utilities.html>The
 Single Unix Specification version 4</a> where applicable. This does not mean
 that Toybox is implementing every SUSv4 utility: some such as SCCS and ed are
 obsolete, while others such as c99 are outside the scope of the project.
 Toybox also isn't implementing full internationalization support: it should be
 8-bit clean and handle UTF-8, but otherwise we leave this to X11 and higher
 layers. And some things (like $CDPATH support in "cd") await a good
 explanation of why to bother with them. (The standard provides an important
 frame of reference, but is not infallable set of commandments to be blindly
 obeyed.)</p>

 <p>The other major sources of commands are the Linux man pages, and testing
 the behavior of existing commands (although not generally looking at their
 source code), including the commands in Android's toolbox. SUSv4 does not
 include many basic commands such as "mount", "init", and "mke2fs", which are
 kind of nice to have.</p>

 <b><h2><a name="downloads" />Download</h2></b>

 <p>This project is maintained as a mercurial archive.  To get a copy of the
 current development version, either use mercurial (hg clone
 http://landley.net/hg/toybox) or click on one of the zip/gz/bz2 links
 at the top of the <a href=/hg/toybox>mercurial archive browser</a> page to get
 an archive of the appropriate version.  Click
 <a href="/hg/toybox?cmd=tags">tags</a> to see all the tagged release
 versions ("tip" is the current development version).</p>

 <p>The maintainer's <a href=/notes.html>development log</a> and the project's
 <a href=http://lists.landley.net/listinfo.cgi/toybox-landley.net>mailing
 list</a> are also good ways to track what's going on with the project.</p>

 <!--
 <b><h2><a name="why">Why do toybox?</h2></b>

 <p>Because smart phones are replacing the PC, and Android must become
 self-hosting to beat the iPhone in establishing the new standard.</p>

 <p>This is the third such major transition in computer history:
 (mainframe-&gt;minicomputer-&gt;microcomputer-&gt;smartphone).
 The mainframe was replaced by the minicomputer, which was replaced by
 the microcomputer (renamed the "personal" computer to make clear you could
 access porn through it), which is being replaced by the smartphone. Nobody
 needed to wait for printouts from a big computer in another building when they
 could use a little one down the hall.  Then nobody needed the big computer
 down the hall when they had a little one on their desk.  Now nobody needs the
 big computer on their desk when they have a little one in their pocket.</p>

 <p>The new platform displaces the old when it becomes natively self hosting.
 Often they leverage existing technology: just as early microcmputers used
 teletypes and televisions for output, phones can use
 <a href=http://us.toshiba.com/accessory/PA3575U-1PRP>USB docking stations</a>
 to access a bigger screen, mouse, keyboard, speakers, etc. Plugging a phone into
 USB even charges the battery. But to use the phone as a development
 workstation, it needs more software, such as a Posix command line, a native
 compiler, and drivers for the USB peripherals.</p>

 <p>The new platform also eventually weans itself off of its dominant language.
 Dalvik is to Android what ROM Basic was to the PC: something it must
 eventually outgrow. Thus toybox is native C code, not Java.</p>

 <b><h3>So why aren't self-hosting smartphones attracting more attention?</h3></b>

 <p>Because most people are focusing on the legacy platforms, not on the new
 stuff. Existing multi-billion dollar industries are getting evicted from their
 decades-old established niche, and are trying to spin the transition as an
 opportunity instead of a forced march onto reservations. When elephants run
 from mice, it's easier to notice the elephants.</p>

 <p>History's our guide here: the previous technology always gets kicked up into
 the "server space", moving from "the thing you stood in front of waiting for
 your printout" to "that thing you sometimes accessed remotely via the new
 computer". This time around they're calling it "the cloud" and pretending it's
 a big deal; it's really just a beowulf cluster with a layer of
 virtualization/containerization software implementing hotplug hardware and
 live migration to provide cheap
 commodity processing power that dominant players (like amazon) literally
 give away for free. These old machines become secondary, only
 accessed through the new machines users now directly interact with.</p>

 <p>Since there's only one server space, the mainframe ate the minicomputer in
 the 1980's (when DEC went under), and this time around "the cloud" seems to be
 eating the mainframe (IBM ain't happy). The inevitable consolidation leads
 to drama, but doesn't mean much in the long run.</p>

 <p><a href=http://landley.net/notes-2012.html#12-07-2012>For more
 on this topic...</a></p>

 <b><h3><a name="why_android">Why is Android important?</h2></b>

 <p>Major hardware transitions introduce
 <a href=http://landley.net/notes-2011.html#26-06-2011>new software
 standards</a> which are extremely sticky once
 established, due to network effects.</p>

 <p>Last time around, the PC was stuck with
 a proprietary operating system (DOS/Windows) which is still dominant on that
 hardware platform's descendants 30 years later. This time around, the choice
 is between Android (a Linux derivative) and iPhone (a closed BSD fork ala
 SunOS, put out by a company already engaged in multiple aggressive IP lawsuits).
 The main difference between Apple and Microsoft is that Apple is competent.</p>

 <p>And yes, it has to be Android, it won't be vanilla Linux any time soon,
 for three reasons. 1) <a href=http://landley.net/notes-2010.html#13-08-2010>Open
 Source can't do user interfaces</a> for about the same reason wikipedia can't
 write a novel, 2) it's too late to the
 party (a 5 year headstart is forever in computers), 3) preinstalls matter
 (GPLv3 spooked all the hardware vendors, Android has a "no GPL in
 userspace" policy which is rigidly enforced).</p>

 <p>And "any time soon" is important: attempting to displace an existing
 entrenched de-facto standard is what linux has spent the last 20
 years trying (and failing) to do on the desktop. Spending another 20
 years fighting for less than 1% of the phone market would just be sad.</p>

 <b><h3><a name="how_google">How is Google less evil than Apple?</h3></b>

 <p>Because Android isn't Google's core business, attaching advertising to large
 scale data searches is. Android and Chrome and such are Google's way of
 "commoditizing their co-factors" to drive down the price of ingredients
 to their core business.</p>

 <p>Thus Google is pursuing a commodity market and encouring as many vendors as
 possible to participate, not to control the new space but to hold it open,
 so that its search products are widely available without requiring the
 permission of some other monopoly gatekeeper. Apple is attemping to corner the
 smartphone market and extract monopoly rents, excluding all
 vendors except itself.</p>

 <p>So if Google wins we get a commodity market in smartphone/tablet software,
 and may be able to open it further in future. If Apple wins we get a proprietary
 smartphone/tablet OS with a single monopoly vendor, which is likely to close it
 further.</p>

 <b><h3>Why not just use BusyBox?</h3></b>

 <p>Android can't. Busybox predates android
 by many years; if they were ever going to ship it they'd have done so by
 now. Android has had a "No GPL in Userspace" policy ever since GPLv3
 came out (before the first Android phone shipped), and they mean it.</p>

 <p>Toybox also has a better design and simpler code. I did both
 and this is the one I enjoy banging on; I tried to contribute a few things
 to busybox and it was like crawling through a thornbush of #ifdefs. Busybox
 development is just no fun anymore.</p>

 -->

 <b><h2><a name="toycans" />What's the toybox logo image?</h2></b>

 <p>It's <a href=toycans-big.jpg>carefully stacked soda cans</a>.  Specifically,
 it's a bunch of the original "Coke Zero" and "Pepsi One" cans, circa 2006,
 stacked to spell out the binary values of the ascii string "Toybox", with
 null terminator at the bottom.  (The big picture's on it's side because
 the camera was held sideways to get a better shot.)</p>

 <p>No, it's not photoshopped, I actually had these cans until a coworker
 who Totally Did Not Get It <sup><font size=-3><a href=http://www.timesys.com>tm</a></font></sup> threw them out one day after I'd gone home,
 thinking they were recycling.  (I still have two of each kind, but
 Pepsi One seems discontinued and Coke Zero switched its can color
 from black to grey, presumably in celebration.  It was fun while it lasted...)</p>

 <!--#include file="footer.html" -->
	<html><head><title>What is toybox?</title>
	<!--#include file="header.html" -->

	<h2><a name="what" />What is ToyBox?</h2>

	<p>Toybox combines the most common Linux command line utilities together into
	a single <a href=license.html>BSD-licensed</a> executable. It's simple, small, fast, and reasonably
	standards-compliant (<a href=http://opengroup.org/onlinepubs/9699919799>POSIX-2008</a> and <a href=http://refspecs.linuxfoundation.org/LSB_4.1.0>LSB 4.1</a>).</p>

	<p>Toybox's 1.0 release goal is to turn generic Android into a
	development environment capable of compiling <a href=http://www.linuxfromscratch.org>Linux From Scratch</a>.
	A tiny system <a href=/aboriginal>built from</a> just toybox, linux, <a href=http://musl-libc.org>a C library</a>, and a C compiler (such as LLVM or
	gcc 4.2.1+binutils 2.17) should be
	able to rebuild itself from source code without needing any other packages.</p>

	<b><h2><a name="status" />What commands are implemented in Toybox?</h2></b>

	<p>The current list of commands implemented by toybox is on the
	<a href=status.html>status page</a>, which is updated each release.
	There is also <a href=roadmap.html>roadmap</a> listing all planned commands for the
	1.0 release.</p>

	<p>In general, configuring toybox for "defconfig" enables all the commands
	compete enough to be useful. Configuring "allyesconfig" enables partially
	implemented commands as well, along with debugging features.</p>

	<p>Several toybox commands can do things other vesions can't. For example
	the toybox "df" isn't confused by initramfs the way other df implementations
	are. (If initramfs is visible, df shows it like any other mount point.)</p>

	<b><h3>Command Shell</h3></b>
	<p>The Toybox Shell (toysh) aims to be a reasonable bash replacement. It
	implements the "sh" and "toysh" commands, plus the built-in commands "cd" and
	"exit". This is the largest single sub-project in toybox.</p>

	<p>The following additional commands may be built into the shell (but not as
	separate executables): cd, exit, if, while, for, function, fg, bg, jobs, source,
	<a href="http://opengroup.org/onlinepubs/9699919799/utilities/alias.html">alias</a>,
	export, set, unset, read, trap, and exec. (Note: not done yet.)</p>

	</ul>

	<h2><a name="commands" />Which commands are planned?</h2>

	<p>The toybox <a href=todo.txt>todo list</a> mentions many potential commands
	which may be added to this project. (Whether that file is readable by anybody
	but the project's maintainer is open to debate.) The roadmap wiki in the
	nav bar has a more human readable version.</p>

	<p>The criteria for a toybox 1.0 release is that a system built from just the
	Linux kernel, toybox, C library (such as uClibc), and a compiler (such as
	tinycc) can rebuild itself from source code.</p>

	<b><h3>Relevant Standards</h3></b>

	<p>Most commands are implemented according to
	<a href=http://opengroup.org/onlinepubs/9699919799/idx/utilities.html>The
	Single Unix Specification version 4</a> where applicable. This does not mean
	that Toybox is implementing every SUSv4 utility: some such as SCCS and ed are
	obsolete, while others such as c99 are outside the scope of the project.
	Toybox also isn't implementing full internationalization support: it should be
	8-bit clean and handle UTF-8, but otherwise we leave this to X11 and higher
	layers. And some things (like $CDPATH support in "cd") await a good
	explanation of why to bother with them. (The standard provides an important
	frame of reference, but is not infallable set of commandments to be blindly
	obeyed.)</p>

	<p>The other major sources of commands are the Linux man pages, and testing
	the behavior of existing commands (although not generally looking at their
	source code), including the commands in Android's toolbox. SUSv4 does not
	include many basic commands such as "mount", "init", and "mke2fs", which are
	kind of nice to have.</p>

	<b><h2><a name="downloads" />Download</h2></b>

	<p>This project is maintained as a mercurial archive. To get a copy of the
	current development version, either use mercurial (hg clone
	http://landley.net/hg/toybox) or click on one of the zip/gz/bz2 links
	at the top of the <a href=/hg/toybox>mercurial archive browser</a> page to get
	an archive of the appropriate version. Click
	<a href="/hg/toybox?cmd=tags">tags</a> to see all the tagged release
	versions ("tip" is the current development version).</p>

	<p>The maintainer's <a href=/notes.html>development log</a> and the project's
	<a href=http://lists.landley.net/listinfo.cgi/toybox-landley.net>mailing
	list</a> are also good ways to track what's going on with the project.</p>

	<!--
	<b><h2><a name="why">Why do toybox?</h2></b>

	<p>Because smart phones are replacing the PC, and Android must become
	self-hosting to beat the iPhone in establishing the new standard.</p>

	<p>This is the third such major transition in computer history:
	(mainframe->minicomputer->microcomputer->smartphone).
	The mainframe was replaced by the minicomputer, which was replaced by
	the microcomputer (renamed the "personal" computer to make clear you could
	access porn through it), which is being replaced by the smartphone. Nobody
	needed to wait for printouts from a big computer in another building when they
	could use a little one down the hall. Then nobody needed the big computer
	down the hall when they had a little one on their desk. Now nobody needs the
	big computer on their desk when they have a little one in their pocket.</p>

	<p>The new platform displaces the old when it becomes natively self hosting.
	Often they leverage existing technology: just as early microcmputers used
	teletypes and televisions for output, phones can use
	<a href=http://us.toshiba.com/accessory/PA3575U-1PRP>USB docking stations</a>
	to access a bigger screen, mouse, keyboard, speakers, etc. Plugging a phone into
	USB even charges the battery. But to use the phone as a development
	workstation, it needs more software, such as a Posix command line, a native
	compiler, and drivers for the USB peripherals.</p>

	<p>The new platform also eventually weans itself off of its dominant language.
	Dalvik is to Android what ROM Basic was to the PC: something it must
	eventually outgrow. Thus toybox is native C code, not Java.</p>

	<b><h3>So why aren't self-hosting smartphones attracting more attention?</h3></b>

	<p>Because most people are focusing on the legacy platforms, not on the new
	stuff. Existing multi-billion dollar industries are getting evicted from their
	decades-old established niche, and are trying to spin the transition as an
	opportunity instead of a forced march onto reservations. When elephants run
	from mice, it's easier to notice the elephants.</p>

	<p>History's our guide here: the previous technology always gets kicked up into
	the "server space", moving from "the thing you stood in front of waiting for
	your printout" to "that thing you sometimes accessed remotely via the new
	computer". This time around they're calling it "the cloud" and pretending it's
	a big deal; it's really just a beowulf cluster with a layer of
	virtualization/containerization software implementing hotplug hardware and
	live migration to provide cheap
	commodity processing power that dominant players (like amazon) literally
	give away for free. These old machines become secondary, only
	accessed through the new machines users now directly interact with.</p>

	<p>Since there's only one server space, the mainframe ate the minicomputer in
	the 1980's (when DEC went under), and this time around "the cloud" seems to be
	eating the mainframe (IBM ain't happy). The inevitable consolidation leads
	to drama, but doesn't mean much in the long run.</p>

	<p><a href=http://landley.net/notes-2012.html#12-07-2012>For more
	on this topic...</a></p>

	<b><h3><a name="why_android">Why is Android important?</h2></b>

	<p>Major hardware transitions introduce
	<a href=http://landley.net/notes-2011.html#26-06-2011>new software
	standards</a> which are extremely sticky once
	established, due to network effects.</p>

	<p>Last time around, the PC was stuck with
	a proprietary operating system (DOS/Windows) which is still dominant on that
	hardware platform's descendants 30 years later. This time around, the choice
	is between Android (a Linux derivative) and iPhone (a closed BSD fork ala
	SunOS, put out by a company already engaged in multiple aggressive IP lawsuits).
	The main difference between Apple and Microsoft is that Apple is competent.</p>

	<p>And yes, it has to be Android, it won't be vanilla Linux any time soon,
	for three reasons. 1) <a href=http://landley.net/notes-2010.html#13-08-2010>Open
	Source can't do user interfaces</a> for about the same reason wikipedia can't
	write a novel, 2) it's too late to the
	party (a 5 year headstart is forever in computers), 3) preinstalls matter
	(GPLv3 spooked all the hardware vendors, Android has a "no GPL in
	userspace" policy which is rigidly enforced).</p>

	<p>And "any time soon" is important: attempting to displace an existing
	entrenched de-facto standard is what linux has spent the last 20
	years trying (and failing) to do on the desktop. Spending another 20
	years fighting for less than 1% of the phone market would just be sad.</p>

	<b><h3><a name="how_google">How is Google less evil than Apple?</h3></b>

	<p>Because Android isn't Google's core business, attaching advertising to large
	scale data searches is. Android and Chrome and such are Google's way of
	"commoditizing their co-factors" to drive down the price of ingredients
	to their core business.</p>

	<p>Thus Google is pursuing a commodity market and encouring as many vendors as
	possible to participate, not to control the new space but to hold it open,
	so that its search products are widely available without requiring the
	permission of some other monopoly gatekeeper. Apple is attemping to corner the
	smartphone market and extract monopoly rents, excluding all
	vendors except itself.</p>

	<p>So if Google wins we get a commodity market in smartphone/tablet software,
	and may be able to open it further in future. If Apple wins we get a proprietary
	smartphone/tablet OS with a single monopoly vendor, which is likely to close it
	further.</p>

	<b><h3>Why not just use BusyBox?</h3></b>

	<p>Android can't. Busybox predates android
	by many years; if they were ever going to ship it they'd have done so by
	now. Android has had a "No GPL in Userspace" policy ever since GPLv3
	came out (before the first Android phone shipped), and they mean it.</p>

	<p>Toybox also has a better design and simpler code. I did both
	and this is the one I enjoy banging on; I tried to contribute a few things
	to busybox and it was like crawling through a thornbush of #ifdefs. Busybox
	development is just no fun anymore.</p>

	-->

	<b><h2><a name="toycans" />What's the toybox logo image?</h2></b>

	<p>It's <a href=toycans-big.jpg>carefully stacked soda cans</a>. Specifically,
	it's a bunch of the original "Coke Zero" and "Pepsi One" cans, circa 2006,
	stacked to spell out the binary values of the ascii string "Toybox", with
	null terminator at the bottom. (The big picture's on it's side because
	the camera was held sideways to get a better shot.)</p>

	<p>No, it's not photoshopped, I actually had these cans until a coworker
	who Totally Did Not Get It <sup><font size=-3><a href=http://www.timesys.com>tm</a></font></sup> threw them out one day after I'd gone home,
	thinking they were recycling. (I still have two of each kind, but
	Pepsi One seems discontinued and Coke Zero switched its can color
	from black to grey, presumably in celebration. It was fun while it lasted...)</p>

	<!--#include file="footer.html" -->