This draft is a start on a Documentation file for new Linux
projects in the family of SubmittingPatches and CodingStyle.
It attempts to capture advice kernel maintainers repeatedly give
to large Linux project developers, especially those new to Linux.
With an increasing amount of software development taking place
in the Linux environment, it is hoped this contributes in some
small way to help people avoid the mistakes of those who have gone
before them in getting their code accepted into the Linux kernel.
Inspired by Andrew Morton's post on lkml just a while back on this
subject.
If including this file in the Documentation directory is agreed to,
I'll be glad to incorporate feedback and resubmit.
Any thoughts?
thanks,
Nivedita
--------------------
File: SuccessfulProjects.txt
Date: 3/14/04
Title: How To Run A Successful Linux Project
" How to improve your chances of launching and sustaining a successful Linux
project, get your code or technology accepted into the Linux kernel and
adopted by the community, earn fame (or employment, or at least continued
employment, or well, at least not completely waste your spare time), all
without losing your hair and your sanity. "
Goal
====
- Increase the success rate of Linux development projects
- Reduce the burden on the kernel maintainers and the community
- Decrease the angst and conflict experienced by project developers
- Make software development faster and more efficient
- Make users, consumers of those software projects happier
Introduction
============
Most of the information here is very basic, obvious and covered frequently in a
multitude of places, at length. However, it is also difficult to locate in one
convenient place, and ignored frequently enough to provoke the presence of this
file in the kernel Documentation subdirectory.
Tips
====
1]. Become familiar with Linux kernel development!
--------------------------------------------------
1.1 Who are the maintainers affected?
Learn who the maintainers are for the subsystems affected by your project,
and for the various releases, especially for the releases you intend to
provide code to.
2.4 -> Marcelo Tosatti
2.6 -> Andrew Morton
development -> Linus
Maintainers file -> current list of maintainers
1.2 Which are the mailing lists you need?
Learn which mailing lists cover development in the areas affected by your
project. It is always a good idea to involve the kernel community or
sub-community as the case may be - which involves posting to the right
mailing lists. Solicit advice on which lists are appropriate.
You can start by checking the MARC archives to find the right lists.
http://marc.theaimsgroup.com/
1.3 Learn Linux Kernel Mailing List (lkml) etiquette
Read the Linux Kernel Mailing List FAQ.
http://www.tux.org/lkml/
1.4 Which Linux source tree?
Learn what a stable release and what a development release is.
Read the Linux FAQ.
http://en.tldp.org/FAQ/Linux-FAQ/index.html
This is, admittedly, already slightly out of date.
Also discover the various other source trees (-mm, -osdl, ...)
Read the Linux Kernel Newbies FAQ.
http://kernelnewbies.org/faq/
Get advice from the community and your users on which kernel tree would be
best to target for inclusion. Understand that kernel development occurs in
parallel in various source trees, and you might need to provide support in
multiple versions.
2]. Join or Create a Community
------------------------------
2.1 Join an existing community
If there already exists a project developing functionality foo that you
are interested in, work with it. Join the people who have already spent
time and effort solving the problem. Sometimes, this is easier said
than done because projects might be open source in name, but far from it in
reality. A good open source project, however, will have a public web site,
a public mailing list that invites discussion and source code available to
play with.
If you cannot convince this community of the value of your ideas, the going
will only get tougher when taken to the linux kernel community. Not always
true, but true often enough.
This is particularly true when APIs have to be designed and there are no
mandated standards controlling what you should be implementing. Having
multiple conflicting implementations brought to the linux kernel mailing
list puts the burden of sorting out basic issues related to your project
on kernel maintainers, hardly a group with spare time on their hands.
2.2 Create an open source community if there is none
If there is no existing project that meets your needs, create one.
Maintain the public infrastructure that should ideally accompany a large
project - a public website, mailing list, source code development
infrastructure (SourceForge is a good place to start). Described well
in the links below.
Read the Software Release Practices Howto.
http://en.tldp.org/HOWTO/Software-Release-Practice-HOWTO/index.html
Read the following documents in the Documentation subdirectory:
CodingStyle
SubmittingPatches
SubmittingDrivers
Do not take any of the above as gospel, confirm with the maintainers in
question. There are exceptions to every rule.
3. Interact early, interact often
---------------------------------
3.1 Don't work in isolation
It is not a good idea to spend several person-years working behind closed
doors, or even within your own project environment. Keep not only the
project community involved, but also the maintainers concerned and the
linux kernel community, if appropriate, in the loop.
In Andrew Morton's words:
"But beware of being *too* disconnected from the [email protected].
We don't want to get in the situation where you pop up with a couple
of person-years' worth of work and other kernel developers have major
issues with it. Please find a balance - some way of regularly
checkpointing."
http://marc.theaimsgroup.com/?l=linux-kernel&m=107922697510704&w=2
Even if you are in design on planning stages, it is worth a note to
the community to say, "Hey, this is how we're going to go about it..."
i.e. remain visible, and ensure that people know your project is alive
and in good hands.
3.2 Avoid large code dumps
Don't throw a massive, complex tarball of your final masterpiece at the
kernel community and maintainers once you are done. Break down your project
into smaller pieces. Submit easily digestible chunks at regular intervals.
If you're making some ghastly, widespread mistake, catch it early. Get
agreement from the community and the maintainers on your approach. Again,
to quote Andrew from the link above:
"That way everyone else can see the code evolving, and can help, and
can understand. And other people will fix your bugs for you, and
update your code as kernel-wide changes are implemented. And we all
avoid nasty surprises and extensive rework."
3.3 Be responsive to input from the community
Good open source project maintainers earn the trust of the larger
community and kernel maintainers by demonstrating they are willing to work
in tandem with the community.
See Greg Kroah-Hartman's slides on dealing with the community.
http://www.kroah.com/linux/talks/cgl_talk_2002_10_16/
Nivedita Singhvi <[email protected]> wrote:
>
> This draft is a start on a Documentation file for new Linux
> projects in the family of SubmittingPatches and CodingStyle.
>
> It attempts to capture advice kernel maintainers repeatedly give
> to large Linux project developers, especially those new to Linux.
> With an increasing amount of software development taking place
> in the Linux environment, it is hoped this contributes in some
> small way to help people avoid the mistakes of those who have gone
> before them in getting their code accepted into the Linux kernel.
> Inspired by Andrew Morton's post on lkml just a while back on this
> subject.
>
> If including this file in the Documentation directory is agreed to,
> I'll be glad to incorporate feedback and resubmit.
>
> Any thoughts?
Looks useful, thanks. I guess I'm not really in the target audience, so
I'll probably miss things.
> --------------------
> File: SuccessfulProjects.txt
> Date: 3/14/04
> Title: How To Run A Successful Linux Project
>
> " How to improve your chances of launching and sustaining a successful Linux
> project, get your code or technology accepted into the Linux kernel and
> adopted by the community, earn fame (or employment, or at least continued
> employment, or well, at least not completely waste your spare time), all
> without losing your hair and your sanity. "
>
> Goal
> ====
> - Increase the success rate of Linux development projects
> - Reduce the burden on the kernel maintainers and the community
> - Decrease the angst and conflict experienced by project developers
> - Make software development faster and more efficient
> - Make users, consumers of those software projects happier
>
- Use the kernel's review processes and testing base to increase the
quality of your software.
>
> Introduction
> ============
> Most of the information here is very basic, obvious and covered frequently in a
> multitude of places, at length. However, it is also difficult to locate in one
> convenient place, and ignored frequently enough to provoke the presence of this
> file in the kernel Documentation subdirectory.
>
>
> Tips
> ====
>
Read Documentation/CodingStyle! If the code doesn't look like kernel code
you've just made things much harder for yourself.
> 1]. Become familiar with Linux kernel development!
> --------------------------------------------------
> 1.1 Who are the maintainers affected?
> Learn who the maintainers are for the subsystems affected by your project,
> and for the various releases, especially for the releases you intend to
> provide code to.
> 2.4 -> Marcelo Tosatti
> 2.6 -> Andrew Morton
> development -> Linus
> Maintainers file -> current list of maintainers
./MAINTAINERS
>
> 1.2 Which are the mailing lists you need?
> Learn which mailing lists cover development in the areas affected by your
> project. It is always a good idea to involve the kernel community or
> sub-community as the case may be - which involves posting to the right
> mailing lists. Solicit advice on which lists are appropriate.
> You can start by checking the MARC archives to find the right lists.
> http://marc.theaimsgroup.com/
Well. Hopefully simply *reading* the mailing lists will help get people up
to speed.
> 1.3 Learn Linux Kernel Mailing List (lkml) etiquette
> Read the Linux Kernel Mailing List FAQ.
> http://www.tux.org/lkml/
>
> 1.4 Which Linux source tree?
> Learn what a stable release and what a development release is.
> Read the Linux FAQ.
> http://en.tldp.org/FAQ/Linux-FAQ/index.html
> This is, admittedly, already slightly out of date.
It would be a good idea to spell this out here. People often do
development against kernels which, frankly, are not development kernels.
Apart from that, heck, why not? Please run up a diff.
Andrew Morton wrote:
> Apart from that, heck, why not? Please run up a diff.
Thanks for the input, Andrew! I'll resubmit a new version
of this by tonight.
many thanks!
Nivedita
Andrew Morton wrote in an earlier thread:
> Apart from that, heck, why not? Please run up a diff.
Here is the new version of the file, incorporating feedback.
Thanks!
Nivedita
---
File: SuccessfulProjects.txt
Date: 3/14/04
Title: How To Run A Successful Linux Project
" How to improve your chances of launching and sustaining
a successful Linux project, get your code or technology
accepted into the Linux kernel and adopted by the
community, earn fame (or employment, or at least continued
employment, or well, at least not completely waste your
spare time), all without losing your hair and your sanity. "
Goal
====
- Increase the success rate of Linux development projects
- Reduce the burden on the kernel maintainers and the community
- Decrease the angst and conflict experienced by project developers
- Make software development faster and more efficient
- Make users, consumers of those software projects happier
- Use the kernel's review processes and testing base to
increase the quality of your software
Introduction
============
Most of the information here is very basic, obvious and covered
frequently in a multitude of places, at length. However, it is also
difficult to locate in one convenient place, and ignored frequently
enough to warrant the presence of this file in the Documentation
subdirectory of the kernel source.
Tips
====
1]. Become familiar with Linux kernel development!
--------------------------------------------------
1.1 Read Documentation/CodingStyle!
If the code doesn't look like kernel code, you've just made things
much harder for yourself.
Read Documentation/SubmittingPatches.
Read Documentation/SubmittingDrivers, if applicable.
1.2 Who are the maintainers affected?
Learn who the maintainers are for the subsystems affected by your
project, and for the various releases, especially for the releases
you intend to provide code to.
2.4 -> Marcelo Tosatti
2.6 -> Andrew Morton
development -> Linus
./Maintainers file -> current list of maintainers
1.3 Which Linux source tree?
Linux kernel versions are of the form:
major.minor.merry_go_round
The major number is only incremented when a new epoch is warranted,
and they come very few and far between. Only one so far since 1.0.
The minor number is the key here. Even numbered minor versions are
"stable" kernel releases, and odd numbered ones are "development"
releases. Typically, new development (especially major new features,
subsystem overhauls, API breakages) is done in the development tree.
The latest (3/18/04) stable kernel version is 2.6, and the next
development version, 2.7, will be forked off 2.6 once it is deemed
ready. Earlier stable versions - 2.2 and 2.4 - continue to be updated.
The "merry_go_round" number is incremented rapidly enough to render
most followers dizzy, a good thing, since it allows for the rapid
delivery of changes and bug fixes.
Read the Linux FAQ
http://en.tldp.org/FAQ/Linux-FAQ/index.html
This is, admittedly, already slightly out of date.
Also discover the various other source trees (-mm, -mjb, -osdl, ...)
Read the Linux Kernel Newbies FAQ
http://kernelnewbies.org/faq/
Get advice from the community and your users on which kernel tree
would be best to target for inclusion. Understand that kernel
updates occur in parallel in various source trees, and you might
need to provide support in multiple versions.
1.4 Which are the mailing lists you need?
Learn which mailing lists cover development in the areas affected by
your project. It is always a good idea to involve the kernel community
or sub-community as the case may be - which involves posting to the
right mailing lists. Solicit advice on which lists are appropriate.
You can start by checking the MARC archives to find the right lists.
http://marc.theaimsgroup.com/
1.5 Learn Linux Kernel Mailing List (lkml) etiquette
Read the Linux Kernel Mailing List FAQ
http://www.tux.org/lkml/
2. Interact early, interact often
---------------------------------
2.1 Don't work in isolation
It is not a good idea to spend several person-years working behind
closed doors, or even within your own project environment. Keep not
only the project community involved, but also the maintainers concerned
and the Linux kernel community, if appropriate, in the loop.
In Andrew Morton's words:
"But beware of being *too* disconnected from the
[email protected]. We don't want to get in
the situation where you pop up with a couple of
person-years' worth of work and other kernel developers
have major issues with it. Please find a balance - some
way of regularly checkpointing."
http://marc.theaimsgroup.com/?l=linux-kernel&m=107922697510704&w=2
Even if you are in design/planning stages, it is worth a note to
the community to say, "Hey, this is how we're going to go about
it..."
i.e. remain visible, and ensure that people know your project is
alive and in good hands.
2.2 Avoid large code dumps
Don't throw a massive, complex tarball of your final implementation
at the kernel community and maintainers once you are done. Break
down your project into smaller pieces. Submit easily digestible
chunks at regular intervals. If you're making some ghastly, widespread
mistake, catch it early. Get agreement from the community and the
maintainers on your approach. Again, to quote Andrew from the link
above:
"That way everyone else can see the code evolving,
and can help, and can understand. And other people
will fix your bugs for you, and update your code as
kernel-wide changes are implemented. And we all
avoid nasty surprises and extensive rework."
2.3 Be responsive to input from the community
Good open source project maintainers earn the trust of the larger
community and kernel maintainers by demonstrating they are willing
to work in tandem with the community.
See Greg Kroah-Hartman's slides on dealing with the community.
http://www.kroah.com/linux/talks/cgl_talk_2002_10_16/
3]. Where to start?
-------------------
3.1 Join an existing community
If there already exists a project developing functionality foo that
you are interested in, work with it. Join the people who have
already spent time and effort solving the problem. Sometimes, this is
easier said than done because projects might be open source in name,
but far from it in reality. A good open source project, however,
will have a public web site, a public mailing list that invites
discussion and source code available to play with.
If you cannot convince this community of the value of your ideas,
the going will only get tougher when taken to the Linux kernel
community. Not always true, but true often enough.
This is particularly true when APIs have to be designed and there
are no mandated standards controlling what you should be implementing.
Having multiple conflicting implementations brought to the Linux
kernel mailing list puts the burden of sorting out basic issues
related to your project on kernel maintainers, hardly a group with
spare time on their hands.
3.2 Create an open source community if there is none
If there is no existing project that meets your needs, create one.
Maintain the public infrastructure that should ideally accompany a large
project - a public website, mailing list, source code development
infrastructure (SourceForge is a good place to start). Described well
in the link below.
Read the Software Release Practices Howto.
http://en.tldp.org/HOWTO/Software-Release-Practice-HOWTO/index.html
Do not take any of the above as gospel, confirm with the maintainers in
question. There are exceptions to every rule!.