Process Improvements in the Kernel Development

Kalki writes "In an e-mail to the Linux kernel mailing list, sent Saturday, Torvalds proposed that kernel developers begin certifying that the code that they contribute is entitled to be included in the Linux kernel as well as a technique for "signing off on patches" that would better track which developers had handled source code contributions. check this Infoworld story on it."

Comments

heheh

[RupertJ]

"..."signing off on patches" that would better track which developers had handled source code..."

.... and Linux joined the world of professional software development!! =)

/me ducks

Re:heheh

[drooling-dog]

.... and Linux joined the world of professional software development!! =)

I hope not, since the "unprofessional" model has worked so well (and I'm not being sarcastic). This is more an acknowledgement that GNU/Linux is swimming in dangerous waters, and has enemies with money to burn. Even though SCO's claims have apparently turned out to be lame, you have to assume that intellectual property traps are being set left and right.

A good thing.

[Raven42rac]

The more organisation and delegation in Linux, the better. With the gains being made using Bitkeeper and this, I feel that Linux will make leaps and bounds in the next year. The things I hope for are better hardware detection and working device drivers for more devices (especially multifunction printers). I think Xandros is getting really close to the way things can be. But then again, I run a Debian CLI install on a Pentium II 350. I guess what I meant to say was, for Linux to gain mass market acceptance, it needs to do everything Windows/OS X does, but better, cheaper and faster.

Linus retiring?

[johnthorensen]

This is very speculative, but this looks like the sort of thing somebody does to ease the transition when they turn something over to another leader. Yeah, it's about a 1:10000 chance that I'm right, but remember you heard it here first...

-JT

Re:Linus retiring?

[xlyz]

This is very speculative, but this looks like the sort of thing somebody does to ease the transition when they turn something over to another leader. Yeah, it's about a 1:10000 chance that I'm right, but remember you heard it here first...

you know what? I agree
in 50 year max Linus will hand over his responsability
please add me to the credit list of the future-tellers

Re:Linus retiring?

[skidv]

Remember that Linus "retires" from a version of the kernel when he thinks it is stable enough for a maintainer to supervise. Once he is freed from maintaining the current "release" version of the kernel, he starts working on the next development version.

He'll pass version 2.6 to someone and then start work on 2.7, just as he passed 2.4 to Marcelo Tosatti and then began working on 2.5.

Re:Linus retiring?

[Anonymous Coward]

He's already handed over much of the release management for 2.6 to Andrew Morton. Linus is more focused on the development tree at the moment.

Re:Linus retiring?

[Turmio]

And that someone would be Andrew Morton.

Good Idea

[Anonymous Coward]

Good idea, even if some people will say that SCO scared them into doing it (they say it is a factor, but not the driving reason)

Accounting

[dimss]

Over years, Linux development team has become an enterprise. Finally they realised that they need accounting.

Re:Accounting

[xlyz]

Over years, Linux development team has become an enterprise. Finally they realised that they need accounting.

now if only the "proprietary" software developer will accept external audit to verify they are not using sources they are not entitled to, we will be all set

Re:Accounting

[swillden]

Over years, Linux development team has become an enterprise. Finally they realised that they need accounting.

Just a clarification: What Linus is doing is making the accountability easier and somewhat more complete, not adding it. As he pointed out in his LKML post, Linux developers have been able to find the origin of every bit of code they've needed to, but the process has been painful and has required a little guesswork, particularly for the oldest stuff.

What he's proposing here is just a slight formalization and elaboration of the process that has been used for years. Currently, if I submit a patch to LKML to fix, say, a VFS bug, it will get poked, prodded and adjusted on the mailing list until people think it's clean and solid. Then the subsystem maintainer (Al Viro, in this case) will pick it up, probably tweak it some more, attach a "From" comment, stating that I am the author and forward it to Linus. Linus will review it, accept it, and his scripts will add my name into the changelog and the CREDITS file.

Since all of this happens on the public, archived, mailing list, there's plenty of accountability, but figuring out the sequence of events requires digging through the archives, and there may not be any obviously ideal search criteria.

Now, Linus wants me to attach my name myself, and to do it in a standardized format so that it's more searchable. Further, he wants everyone else who modifies the patch in any way to add their stamp as well, providing a change history in the patch itself. It's a weak change history, since it doesn't describe what changed, but it provides the starting point for searching the archives.

So, what Linus is asking for isn't so much to create a better accountability trail as it is to make the existing trail easier to follow. It's an ease-of-use optimization.

Well, there is one way in which this is perhaps a significant enhancement, and that is that Linus wants to formally define the legal commitment a contributor makes. In a reasonable world, this should be unnecessary, since if I contribute some code that I don't own, I should be the one held liable for the copyright infringement, not the others who used it in good faith. In the litigious world we live in, however, it's a good idea to formally spell it out, and make clear to everyone that by attaching their name to a patch, they're providing a certain warranty of their right to contribute it.

Groklaw article

[jadel]

There is an article on this subject at groklaw
It covers more or less the same territory in a bit more depth.

A Good Thing(tm)

[Alizarin Erythrosin]

Anything that prevents the possibility of another SCO-type BS lawsuit is a Good Thing.

Hopefully it can avoid patent issues too. If something goes into Linux and later some company (Microsoft?) files a patent lawsuit, there may be evidence of prior art if the code was "certified" on a certain date.

On the reverse side, it can provide exactly who contributed the code (which can already be done mind you), but this time, they certified it for use, which can possibly cause more legal troubles.

Re:A Good Thing(tm)

[Spoing]

1. Anything that prevents the possibility of another SCO-type BS lawsuit is a Good Thing.

Prevent? Nothing will do that short of turning SCO corp. into a legal crater as an example of any other company foolish enough to attempt the same stunt.

The kernel development process is about as public, auditable, and open as I can imagine. (Yes, there are some with better audit controls, though most of those projects are quite private or secret!)

Any code that has substance goes in has a name associated with it in the change log, and historic versions of the kernel are available going back 10+ years. I don't know of any closed commercial application I've ever worked with that can say the same...usually, it's a couple years, major releases (maybe), and forks. That's it!

tracking

[colinleroy]

"signing off on patches" that would better track which developers had handled source code contributions.

Linus Torvalds' problem is the fact that, as it is currently easy to find out who commited the patch, and often who provided it (which often appears in Bitkeeper's changelog), the whole submission process can be a blackbox - if I send a patch to alsa subsystem's maintainer, he'll probably apply it to alsa's CVS, maybe someone else will modify this patch, and when included in linux' main tree, only the merge information would appear.

Its official then

[Timesprout]

Patches submitted as AC will no longer be included in the Linux kernel

Re:Its official then

[Anonymous Coward]

Patches submitted as AC will no longer be included in the Linux kernel

Damn, bad news for Alan Cox ...

Avoid even the appearance..

[hot_Karls_bad_cavern]

As taught to almost all people taking the moral high-ground, that are in the public eye:

Avoid even the appearance of wrong doing.

Guys, this is a great idea for accountability in the kernel source. In the next round, the "SCO" of that round might not be so blatently stupid and far more sinister. Please watch your code and keep it clean!

Re:Avoid even the appearance..

[BinLadenMyHero]

I wonder if this is in any way related to this.

From FSF:
We have just begun a project here at FSF to document and codify our process, so that it can be disseminated in the form of a policy manual and accompanying software, to all other Free Software projects who wish to solidify their legal assembly process. Distilling nearly two decades of organizational know-how into easy-to-understand software and documentation is no easy task, and we will rely greatly on your financial support to aid us in carrying out this momentous task.

Professional Approach coming to Linux ?

[kbsingh]

Sounds pretty good. I think Linux needs a basic system of this sort in place as-soon-as-practical. It will bring together a lot of accountability of / for code in the Kernel itself. Plus, it should counter any issue like the SCO created one, in the future.

Another interesting point here seems to be that with this management overhead and the admin work that issue such as this create, how much of time is Linus actually spending with them ? while he might be working with the technical side of things ?

Inspite of all the noise, there are just a handfull of people contributing major code into the Kernel ( would 300 be a fair guess ? ) How are all these admin overheads going to effect their performance ? Also is anyone / everyone expected to research the piles and piles of patenets / copyrights before they make such a declaration ?

This quote says it all

[Henrik S. Hansen]

I think this quote really says it all about why this is a good idea:

"People who don't understand how I interact with the people I work with literally feel better just having it down more as a documented process," he [Linus] said.

Re:This quote says it all

[gosand]

I think this quote really says it all about why this is a good idea: "People who don't understand how I interact with the people I work with literally feel better just having it down more as a documented process," he [Linus] said.

Take my comments with a grain of salt, because I am up to my eyes in process development because we are trying to get CMM Level 2 certification where I work.

I don't see a problem at all with documenting the way things are done. I know a lot of people resist it, but think about it. How hard would it be for Linus to just write down how he does things. You'd be surprised how many times you uncover problems (or potential problems) when you have to write down your processes. Sometimes, you immediately see ways to improve things. If not, then at most you are out a little bit of effort.

But I really think that Linus wants to do this so that when he is on the stand and a SCO attorney asks him how code is added to the kernel, he can just say "RTFM!".

[RFD] Explicitly documenting patch submission

[Anonymous Coward]

This is the e-mail itself, as posted to LKML by Linus - on Sunday, not Saturday.

Posted anonymously to avoid karma whoring.

Good Idea

[Whitecloud]

This sounds like a good way to ensure accountability on who made what changes, and when they did it. Linus says the SCO debacle "have provided a "big impetus" for the changes", this will make sure similar legal action can be shot down immediately.

Considering all the code thats been leaked lately, this is a welcome insurance policy to keep Linux on track as free alternative OS.

Link to Linus' Mailing List Post

[sgarrity]

Here's the post discussed in the article.

moderation of patches

[millahtime]

maybe they can put some kind of mod system in for the patches. like /. then those who add patches that aren't coded well could get moded down and have bad karma.

Details?

[gpinzone]

The story is a little light on the details. The change management system they use (I remember hearing it was BitKeeper) ought to log the users who contributed the code automatically. What I'm wondering is if Linus is talking about validating the identity of the contributers. If so, this will prevent any anonymous code contribution. I wonder how many people who work for companies that signed NDAs are contributing code that will no longer be able to do so?

Why do I get the feeling....

[10Ghz]

That we will soon see SCO/AdTI press-releases saying "we were right! There are serious problems with the way Linux-hackers handle the code! After all, if there are no problems, why are they taking these steps to correct the situation? This proves once and for all that our claims regarding Linux are true!"

Re:Why do I get the feeling....

[linuxdoctor]

The problem is that SCO, Microsoft or any other company have even worse problems.

Unless and until a company has in place a development process that conforms to independent and internationally recognized standards such as ISO-9000 and has been certified as such you have no guarantee that what they are doing conforms to good engineering practices.

The truth is Linux development has always been open. SCO and other private companies keep their development process secret. Who knows what they are hiding behind all that secrecy.

For my money, I'd like to see Linux development conform to ISO-9000.

Thread on kerneltrap

[lazy_arabica]

You may read the lkml thread and Linus post on kerneltrap.
Just thought it could be interesting...

His aim is different

[lazy_arabica]

Taken from Linus post (talking about SCO claims) :

People have been pretty good (understatement of the year) at debunking those claims, but the fact is that part of that debunking involved searching kernel mailing list archives from 1992 etc. Not much fun.

I suppose having lawyers read your private mail in order to find some clue about the origine of a piece of code make you a much more careful man.

No Anonymous Code

[Short Circuit]

So what happens to people who want to contribute code, but don't want their name attached to it, for various reasons?

• Such as encryption development in France or China, where unauthorized encryption is illegal, IIRC.
  
• Or some employee whose boss wants to own all his creative work, on and off the clock.
  
• Or people who simply don't want to take the risk of being unfairly targeted by some software company for writing code that looks vaguely like the company's.
  
• Or people who had a great idea, but couldn't possibly know someone else had come up with the idea and copyrighted or patented it.
  
  
  IMO, it has its ups and its downs. It allows a greater degree of delegate-the-blame (Good for any large project, Objectively speaking), but it will reduce contributions.

Re:No Anonymous Code

[sweede]

This is the whole idea of the accountability thing.

If you live in china or where ever and dont want to get in trouble for writing encryption code, DONT. I mean how hard is that? If you choose to do something illegal you SHOULD be accountable for your actions and any repercussions from those actions. You probably bitch and moan on how you where going just "5mph" over the speed limit when you get pulled over for doing 50 in a 30 or complain that you were going around the block and didnt need a seatbelt.

If you sign a contract for work that says your employer owns all of the work you do during non-work hours, you should of read it first. If you did and you signed it anyways, dont bitch about having to give up everything you write.

If a closed source company tries to sue you for thinking that your code is close to theirs, you must ask yourself, how much water does their claim hold if there is no way you can view the sources? in a court case, you, the defense has a right to see their evidence against you, and the code that you are infringing on. You do have rights you know. This is easily solved by saying "whoops, I didnt know and i'll change/remove, show me which lines". Again, this is taking responsibility for your own actions. why do people think they can do things and not take any responsability for it? Worse yet, what if you where the project maintainer? since no one signed the code and now you submitted it, your name is on it and you are in trouble.

this last one is just stupid,
"Or people who had a great idea, but couldn't possibly know someone else had come up with the idea and copyrighted or patented it."

What about having a great idea, not doing jack about it, then 3 years later some company does the same thing and makes a mint from it? What can you do about it then?
you: "hey that was my idea 3 years ago and your violating GPL"
them: "oh ya, wheres your proof"
you: "oh right, i have none"

If you have a great idea that might or might not be copyrighted or patended and your too lazy to search around to see if it actually is, you shouldn't be contributing code to any project.

I guess the only downside that i can think of is that it holds people to take responsibility for what they do.

damn, I forget that's horrible!

The original thread

[42forty-two42]

Here's the thread in question.

Authentication Process

[Master Eclipse]

The authentication needs to be done using GPG (GNU Privacy Guard) or PGP (Pretty Good Privacy). This will prevent anyone in the future from inappropriately placing code in the kernel. These two programs provide an excellent means of determining the authenticity of the author. Moreover, the origins of all code submissions can easily be tracked and catalogued using some open source software some friend of mine and I have been working on. In a nutshell it works like this: Code is received either by FTP, E-mail or (virtually) any other mans. At the end of the encrypted code, the code is signed and encrypted with the writer's key. Each of these keys is kept in a database that contains verified information about the writer. This can include their name, and address, or whatever is appropriate. This database is kept as a public record of which code belongs to whom, and when it was created (or submitted). Think about it... anybody who wants to submit code should not be able to do so anonymously. This stands to reason in light of what has been going on lately with SCO. Moreover, this method looks good to executives who have no idea how software is developed and is a legitimate method of proof. So far as being on the Internet, this project is not right now, some friends of mine and I at the University have been beta testing it and it works wonderfully and is very secure. Thank you for your interest!!!! Any thoughts?

Just what I've been on about!

[Paul Fernhout]

And it did not attract much interest back then. Maybe now is the time for such a peer-to-peer repository tool with related free licensing affirmations. 2001 posting to gnu.misc.discuss and further: second 2001 posting to gnu.misc.discuss

Excerpt from the first: So anyway, what do people think? Would such a license management approach with tools, protocols, and standards (ensuring every work received or sent comes with a legally binding machine-readable license and related audit trails for derived works) promote more clearly titled (in a legal sense) free software or would it be the slippery slope to the "right-to-read" future? Without explicit licensing handling, are we just setting the free software or open content communities up for legal challenges down the road as people just try to do their best without such tools? Is trying to automate license handling really that much different (in a bad way) from the current situation of often distributing zip files including both content and license?

Excerpt from the second: This need for a license is especially true for making derived works you plan to redistribute, because here your liability seems highest. If we are to have a lot of free software and free content, based on fine grain collaboration made possible by the internet allowing us to rapidly modify each others works, that is a lot of licenses citing a lot of authors. If these licenses get lost (as in the midi file download example I cited), the content can no longer be considered free. Handling lots of papers is what computers are good at. The less time people need to spend thinking about, negotiating, and managing paperwork and extra files related to free licenses, the more time they can spend making free software and free content.

When?

[miffo.swe]

Now that the Linux development is totally transparent when will we be able to audit Microsoft, SCO, and other propriarity code for stolen bits and pieces?

We should shout out of the top of our lungs that the propriarity way foster code stealing because no one can audit it.

Re:When?

[vijaya_chandra]

when MS decides to replace kernel.dll with vmlinuz-2.x.x-xx

Do "professional" organizations do this?

[ron_ivi]

"Linux joined the world of professional software development!!"

Do companies like Microsoft or Sun make Developers certify that the code they submit in a particular check-in isn't "borrowed" from GPL'd or other open source stuff?

Seems Linux is well ahead of professional organizations in this respect.