Slashdot Mirror
Linus Tries Out BitKeeper
Flammon writes: "Linus has been overloaded
with patches for a while and recently the issue started to become hot again. In an unprecedented move, Linus has started using BitKeeper, as reported by Linux Today. The benefits of BitKeeper are already showing from the large amount of detail provided in the latest unstable kernel pre-release." eirikref adds: "Read Linus' own statement and take a look at the BK web interface."
IIRC, the PPC Kernel is maintained through BitKeeper, and has been for quite some time.
------
Random, useless fact: I type in startx entirely with my left hand.
I wonder if the nice people at Ask Jeeves are going to mind having their (presumably trademarked) logo swiped for this?
I believe Linus was pretty against CVS from day one. He didn't like it at all, and wasn't a terribly huge fan of BitKeeper either. It almost seems like he is using moreso because he has been prodded in all directions regarding this.
Isn't BitKeeper a (gasp!) closed-source commercial software?
... TUX, Episode I: The revenge of the Borg!!
Shock! Horror! Has Linus Torvalds turned to the dark side of the code?!?!
Stay tuned for the next episode of
The right to offend is far more important than the right not to be offended. (Rowan Atkinson)
Methinks that it wouldn't be trivial to continue to pipe the results in a terse format as well.
For L-K and releases a terse format is appropriate, but I think that keeping the longer ones around somewhere can help some of us newbies understand what the heck is going on in there.
Wouldn't you think that, in the 10 or so years he's been maintaining the kernel, he already evaluated it? Just because there was never a press release doesn't mean he rejected CVS out of hand and has never tinkered with it in his spare time.
It's high time he told the community to screw off for a bit.
It's his friggin' hobby, after all. If people don't like the way he deals with it, maybe they ought to go work for a more personable coder on another OS, like, say, Theo De Raadt.
Scary thought, hey?
--------
Bleah! Heh heh heh... BLEAH BLEAH!!! Ha ha ha ha...
Why would somebody choose BitKeeper over Perforce? Perforce offers free licenses for Open Source software and is IMHO 1000% better and more powerful than CVS. Anybody wants to clarify what makes BitKeeper the tool of choice?
Now I'm confused!
I've been using CVS for years and read with great interest the recent Linux Journal article about the Subversion project to created a CVS replacement that is better than CVS.
Then I see a Slashdot story about arch.
Now, my FearLessLeader starts using Bitkeeper.
Should I move from CVS and, if so, which is best?
"Provided by the management for your protection."
Can it really be a bad thing to have too much information about any changes?
How much time does Linus have to dedicate to all these patches that get submitted ?
Would a seperate fork, with sections maintained by indiviudal groups be best ? 4 or 5 guys in charge of VM, 4 or 5 guys in charge of Hardware, they would only be responsible for review and merging.
I know ill get blasted for fork speak, but sometimes its a good thing, (While youre at it optimize for the x86...lol)
Linus is the all benevolent creator and Linux god granted, respect is due, We however are the users, the ones in need, Linux was intended to fill this need, if it reaches a point because of whatever reason, perhaps a branching, is best for it as a whole. I dont think anyone actually asked Linus if he wanted the development to consume his life (Maybe he does, I dont know, it dosent matter)
All this is an awful lot to ask of any one man, mortal or not. Perhaps Linus would welcome this as an oppurtunity to do other things.
I hope this will make Linus's life easier, Sometimes people continue on a path out of a feeling of obligation, does Linus do this now because, 1 He wants to, 2 He feels like he has to
3 Nobody else has stepped up to offer a solution.
Sig went tro...aahemmm.....fishing........
Bitkeeper is available under two licenses. The commercial license costs money and comes with support. The non-commercial license does not cost money., but it has a requirement that all your ChangeLogs must be sent to a world-readable server controlled by BitMover.
Bitkeeper source is available, but it's illegal to redistribute a version of Bitkeeper with the mandatory open logging stripped out.
Bitmover Inc. wants to avoid the situation where people use bitkeeper like gcc, taking free software tools but not giving anything back. You can pay Bitmover money, or you can use a free-as-in-beer version that is suitable for software libre and unsuitable for closed-source software.
I use CVS all the time, but I know its limitations. Linus was right when he decided not to use CVS, it simply is not reliable enough. But don't blame CVS, it is a good and useful tool; but every tool has its safe zone of "recommended use", and Linux kernel is way beyond that. I say, any project above 50 KLOCs and with 100 revisions on average would be pushing the limits.
-- TeknoHog (Pine.LNX.4.44)
Escher was the first MC and Giger invented the HR department.
Well, at least he runs Linux...
http://bitkeeper.com/Products.Comparisons.Perforce .html
Allthough this is marketing poop so it should be taken with a fine grain of salt, it might answer your question.
Never underestimate the relief of true separation of Religion and State.
In other news, demons all over Hell were seen lacing up their skates for the upcoming hockey match against the U.S. National team.
That is fine; but the most important problems would be absence of changesets (so you can't undo related groups of patches), and absence of tiered repositories (everyone goes to the same, single, central CVS server). It all can work, and it does work as we know, but the more code you write the more difficult the maintenance becomes. Like it or not, CVS is an old software, unchanged for years and full of kludges, and BK is one of new designs.
I don't follow it (that closely) either but it seems like you're talking about 2 different, though related, things.
There's the process of patch submissiong which is pretty much the same - send them to the maintainers, who are supposed to send them to the 'trusted few', who send them to Linus. Some of the recent discussion seems to be over the fact that the maintainers thought they had a direct line to Linus, whilst he didn't see it that way.
Then there's the tool Linus uses to organize his source tree(s). This will allow Linus to speed up the process of applying and testing patches. It will not change the protocol for submitting patches. Maybe he'll give the trusted ones write permissions if they even decide to adopt BK too.
At least that's my impression - I could be, and usually am, wrong :)
So soon we'll start hearing that BitKeeper doesn't scale, right?8-)
Yesterday was the time to do it right. Are we having a REVOLUTION yet?
Hopefully they don't keep the repository on the same machine that hosts their website...
/.'ed the kernel!
We may have
I must say, I *LIKE* all the detail in the changelog now. For a LONG time, I've thought the changelogs for linux have been too understated.
... X hardware, X version, etc"
'More bug fixes for PCMCIA' or 'Patches for USB'. Doesnt really tell me if theres any hope a particular problem I am experiencing with either has been fixed -- nor does it tell me why something that used to work no longer works, and how to re-enable the 'old style' code -- or where I should look for the diff to say to the author "This used to work, since this change, it doesnt anymore
More detail means, for example, I can see from the changelog, when the USB sleep (ie. usb does not come back online automatically when you put your computer sleep, you must either do some fancy footwork beforehand (which doesnt always work), or reboot). Its a known problem, but "More USB bugfixes" doesnt tell me its fixed, or even that that part of the code has been worked on.
I'm sure theres many others out there who experience problems in specific parts of the code, (which are known problems), and have been frustrated by the changelog's lack of detail -- and dont want to upgrade your kernel to 2.5, or 2.4-pre's or even another stable 2.4, unless you know your problem is fixed, because what you got now works for everything ELSE, and you never know what a new kernel will break. I myself havnt started using 2.5 kernels, but I would probably start IF I could tell by the changelog, that my problem was solved there, so there was some benefit for me.
Signing up for the mailing list in order to gain access to the bitkeeper download can be a bitch.
The Free Software movement says, "Use the software that's the most free. If you still have a choice, use the best software."
The Open Source Software movement says, "Use the best software. That will often be Free / Open Source software."
Stupid job ads, weird spam, occasional insight at
I am very thankful that Linus finally "saw the light" and started using a source code control system.
:)
I really like the new change logs, I have always hated the old change logs as being too uninformative. One of the really interesting things for me about a source code control system is that it preserves a lot more of the history of the source code than the tar balls do.
It is also really cool how it branches the source for every patch and checks in the code with the users name as the one who checked it in and the body of the email as the comment. If Linus can find a way to also check in his rejected comments on a patch then that will also be very useful. It would be interesting to capture a little bit of the why instead of just the how in the kernel development process.
To apply a patch you just have to merge the branch that contains the patch back into the main development branch, fix any conflicts, compile, fix it so it works right and then commit.
And Linus will never lose another patch again, they will be saved for all time in the source tree under a seperate branch.
Once Linux lets his inner sanctum of kernel developers all start merging approved patches into his main branch then we will see the kernel development really speed up.
Thanks!
-- Never make a general statement.
Maybe they did it to force you to decide if you want to be part of the Free Software/Open Source community or not. It's annoying to me as well when I can't have my cake and eat it too, but I don't complain about the people that make cakes :)
It's just like the GPL - the license is a means to form and defend the community for the good of the community. If you don't like it, that's OK too, but don't say that those community-maintaining features are the problem. They're a feature, not a bug.
Your right to not believe: Americans United for Separation of Church and
Drivers are distributed with the kernel for two reasons:
The USB drivers aren't overly entangled with the real innards of the kernel, they just happen to be shipped in the same tarball.
Your right to not believe: Americans United for Separation of Church and
There are lots of ways of providing such hooks. Perhaps the most compatible with the Linux kernel mindset would be something similar to Emacs-hooks: replace most kernel functions with variables holding function pointers to the actual code and provide APIs for manipulating those hooks.
Some of the newer 'out-of-tree' development that is going on in FreeBSD at the moment is being done using Perforce.
I think some of the SMPng and KSE work is in p4, for example.
I found very interesting a document from Jack Moffitt (of xiph.org fame,
u blic/critique.html
l
one of the main Ogg developers and one of the Icecast Core Developers),
about some problems he had with the BK license when he was using it
for hosting Icecast:
"A Critique of the BitKeeper License"
http://www.mit.edu/afs/athena/user/x/i/xiphmont/P
You might also find interesting his post on the matter to the
"Icecast Developer Discussion List":
http://www.xiph.org/archives/icecast-dev/0067.htm
I hope that he will post here his his experience using BK
in an Open/Free-source project...
Best regards
\\Uriel
P.S.: Yea, I know I'm karma whoring, but I'm sure many people will find this interesting,
specially in casse Jack dont post to this history latter
"When in doubt, use brute force." Ken Thompson
/* you are not expected to understand this */
Potato chips are a by-yourself food.
How does Linus using BitKeeper make kernel development harder for those who don't have it (can't afford, won't use non-free software, whatever)?
Linus and the maintainers will still accept patches in email, so nothing's changed except Linus now has a tool that is likely to help him keep up the extremely high productivity.
And, using non-free software to manage the development of free software doesn't make the free software any less free. It's not like it could only be compiled by a non-free compiler.
Maybe this means that those who write free software will next write a tool even better than BitKeeper and the world will be once more a little better place.
I did a superficial investigation on source control systems, and found some very interesting really free ones, like Aægis.
Does someone know if free alternatives to BK were considered, and if so why a semi-free one was choosen? If BK was better, specifically how it compared to Aægis and other alternatives?
Leandro Guimarães Faria Corcete DUTRA
DA, DBA, SysAdmin, Data Modeller
GNU Project, Debian GNU/Lin
I have yet to see a major new device class, file system class, or other subsystem that didn't require patching. That's a problem with the Linux kernel--it simply lacks the hooks and mechanisms for doing this. And it will only get addressed if the kernel developers start making a commitment to shipping drivers and other modules separately from the main kernel, with their own version numbers and source trees. As long as people can patch easily, they are never going to add the hooks to the kernel that will let new functionality get added without patching.
That is bullsh*it!
The CVS limitation does not lie in how many KLOC or number of revisions it can handle. It handles lot more than that.
The limitations is purely functional, like how it handles branches, merging and such stuff. There it lags behind the newer systems, like Bitkeeper, arch and others.
http://www.millnet.se/ GO/U d- s+:+ a C++ UL++++ P- L+++ E W+++ N+ w++ M-- PE+ t+ X++
> It's somewhat sad that Linux,
Why? I find it interesting.
There's is absolutely NOTHING wrong with charging for software. If you do nothing but write software for work, you have a reasonable expectation to make a living off it. The world doesn't run off charity man, nothing is free.
To me, the "pearl of Free Software" being version controlled by a commercial product is a grand statement.. that free software and commercial software can coexist peacefully.
Software "should" only be free as in speech anyways. If it's simultaneously free beer that's just icing on the cake.
No not "all are heard" in a cell model. It is essentially a partitioning that makes groups more personable, more anarchic and cosy, but does not scale. Think about it: you talk about cells coming together in a larger community so all can be heard. That's fine for smaller groups, but what happens when a church grows to, say, a million or a billion in number? Linux development is more so: communication is intense and data-heavy. The model, as it is now and continues to be, is the traditional hierarchy. Linus trusts a select few, who maintain their own hierarchies. The classic hierarchy is the only structure that works for large groups that need strong direction: armies, nation-states, and yes, some churches.
Well yeah, that's what I took the guy to mean... he never said CVS could not handle that much code functionally, just that it was not suited to projects of that size.
No charge if you let it send your change logs back to bitkeeper for public display, and you get the source, under basically the GPL, with the one restriction that it pass the regression tests (essentailly, you have to let it send the change logs back).
OK, so that's not technically free software, but isn't that close enough?
Plato seems wrong to me today
From Linus' email, mentioned above:
Basically, I'm aiming to be able to accept patches directly from email,
Does Linus use PGP sigs (for example) to verify the senders of these patches? I hope he does (being Linus and all that).
The argument that some people will use is
to say "See? Free software isn't viable on it own. The only reason it's any good at all is that is relies on commercial software"
Or the more subtle "Sure, Linux was okay before. But it only got good once they started using commercial software to develop it"
It could help reinforce the stereotype of free software as "hobby" projects - "Oh sure, you can use free software tools to develop some simple CGI script or napster clone, but if you want to make a serious software project, you need to use commercial tools"
(Not that I believe this, but that is what might be argued)
This is because Linux has a macrokernel architecture - everything's compiled into "the kernel", which is a hassle for some people, but increases execution speed.
As I understand it, WinNT uses a microkernel architecture - the kernel proper does the bare minimum it can get away with, and the rest is handled by higher-level "services", which in theory can be worked on and upgraded without disturbing the microkernel.
Actually, Linux is somewhere between these, owing to the modules system. I agree though that it would be nice if modules were so reliant of what version of the kernel you were using. I don't know about the practicalities of this.
I'll let that self-referential statement stand on its own merits!
Look, if I spend all of my time coming up with "ideas", I have to use them as my source of income; if I can't, I have to find some other source of income, which means I'm not spending any time coming up with those ideas anymore.
Logical. No argument with that. So you can be compensated for the time that you spend doing that through an income that is tied to your production of freely re-distributable ideas. That's the academic model. You work on producing ideas and get a secure, stable life with good benefits.
Or, you could be employed by a company like RedHat. They make money from support/service and use a portion of that to give jobs to talented hackers.
Or, working as a professional software development you can give away some of your time because it increases your standing in the community, raises your profile and in turn attracts employers who wish to pay you for proprietary, non-Free work.
In other words, for decent software to be developed at anything other than a glacial pace, people have to be able to devote time to it, and they have to be compensated for that time in a way that allows them to eat, live under a roof, etc.
In other words, there are several ways in which Free Software is already being developed and in which the developers quite rightly get the food, housing, etc. that you so rightly assert they have the right to be rewarded with. So what's the argument? I would suspect that it is that you are tilting against some fantasy windmill of communist free software.
No flame intended on my side. I just think that you don't get it and that you're arguing against some pre-conceived imaginary position.
The size of the code is a pretty good estimate how many people work on it, how many branches they maintain, and how much of merging they do. So if you know how large your code is, you can tell whether this or that revision control system (or any other tool, to that matter) is appropriate for the job.
No,
if you read the recent thread on l-k, it's because in private Linus has been talking for quite a while to the bitkeeper people about what he wanted from bitkeeper before he'd use it, and the bitkeeper people have gone and implemented most of it, so Linus agreed to use it for a while.
I use Friend/Foe + mod-point modifiers as a karma/reputation system.
I also used to think like this. The sum extent of my source control was cp -r currrent vX.x. Source control was for wimps.
I'm of a rather different view today. I now utterly insist on using it, even in tiny little things that I think are one-offs at the time (quite often it turns out they aren't).
I think I can understand Linus' dislike. It sounds like you're less free, and as if the whole coding thing is suddenly less enjoyable. However, having gone through exactly the same feelings I can say that in my case it certainly isn't true that things are less enjoyable. In fact, in some ways it's easier as I can go wandering off in my own direction for a while, before hitting a dead end and backtracking safe in the knowledge that I have a defined state to fall back on should I need to.
Personally, I'd recommend taking the plunge. Some systems are better than others, but any system use injects a bit more organisation and confidence into the process of coding.
Cheers,
Ian
Sorry about the headline, I couldn't think of a short-enough one that'd fit.
BitMover also has a clause to help free software developers in the event of the company going out of business, located here (Correct me if I am wrong).
It says, '5. CONVERSION TO THE GPL
The BitKeeper Software will be made available under the terms of the GPL in the event that all Open Logging servers cease to function for a continuous period of 180 days starting on or after June 1st 2000.'
This looks to be applicable to the free-as-in-beer version only, as far as I can tell, but still, it's not exactly a big problem.
========================================
Death will come, and will have your eyes
-- Pavese
Will there be public read-only access to Linus' branch so people can keep up with the latest?
You had me at "dicks fuck assholes".
Linus has finally moved from chaos to order. One of the major complaints about the Linux VS BSD development model (a lack of a control system) has been fixed.
Now, to address what this means for Bitkeeper.....its death. Yes. Bitkeeper is now doomed. Why? Simple. The "keep this in the GPL family" movement will have someone clone the Bitkeeper method of software management, and a GPLed Bitkeeper clone will be created, it will catch up to Bitkeeper, pass it, and then Bitkeeper will have its oxygen cut off, and they will die.
Tell that to FreeBSD, OpenBSD, NetBSD, XFree86,
all of which are orders of magnitude larger than the linux kernel. All of them have been using CVS for the past 8-10 years (depending on how you count things). Sure, cvs has its limitations, but the Linux kernel with its small number of developers with write access isn't pushing the limits. FreeBSD has over 250 people committing to its tree right now, for example.
While I'm automatically suspicious of phrases like "Most sane people," I must say both of you have a point.
I disagree that selling software is "software hoarding." Not giving away source code is software hoarding.
Sometimes I think that those of us who advocate Free Software overestimate its revolutionary credentials. I'm of the opinion that Free Software is supported by the most conservative principles that underlie intellectual property law: The creator owns the work and has the right to decide what to do with it and how it is used.
The problem with software is that compilation is tantamount to encryption. When you buy "closed-source" software, you are buying an encrypted book. That doesn't alter the author's right to restrict your rights to copy the work.
The GPL is one collection of grants and reservations of the author's inherent ownership. Other sets are equally valid and based in the same fundamental principle. Where I do agree about "software hoarding" is in the obfuscation of compilation. This creates an artificial shortage of technique.
This cuts the other way, too. When the author of a book steals a paragaph or chapter from another's book, the theft is obvious. When a software author steals a part of another author's program, thanks to the obfuscation of compilation, the theft is not obvious.
What I find funny is the notion that computer software is somehow different. It is not. Copyright, patent, plagarism. These should work the same as in any other form of publishing (I think programming is a form of literature, not a form of invention -- that's my opinion, I think this is a core question not yet settled in public opinion or law).
It is the obfuscation of compilation (provable loss of information) that makes it complex. This is why I think Stallman's definition of "Free Software" is the right one. If you got source code with everything, even if you don't agree with Stallman's particular set of grants and reservations, theft would be easily seen and thus easily prosecuted.
Software vendors think keeping the source secret protects their investment. IMHO it merely drives theft underground. Source always would make it obvious.
Questions: Are there any lawyers in the house? If I write a computer book in French, and one of the chapters is directly translated from someone else's book written in English, am I in violation of copyright? Is it harder to prosecute?
No. It is the nature of big, monolithic programs that don't use modern abstraction facilities to require patches all over. Kernels are no different from word processors in that regard.
That said, your suggestion of using hooks will also require patches 'all over the kernel'.
No. You put in the hooks first, everywhere. Do it from 2.5.10 to 2.5.11. Afterwards, any module can hook just about any function anywhere in the kernel without patches.
Using dynamically loadable modules doesn't make Linux a microkernel, and adding hooks wouldn't either. The performance penalty would be basically zero.
I used this for a couple of years at my last job and it's an excellent product. I used to hate source control systems - especially CVS, but Bitkeeper works excellently.
So DOS the heel out off those servers and make it gpl software!
I wonder who modded that down...
Thats actually an interesting point: the intent of their license is to free the code in the case they go out of business, but the wording leaves open the possibility that an attack could make the source code GPL.
They may wish to consider modifying their license to exempt that case. (They modify the terms so frequently its not a big deal for them) Its not far fetched to imagine a ddos attack run for a few days may cause their service provider to drop them. Combine that with a bit of DNS cache poisoning, and 180 days really isnt that long.
Anyway- the license itself seems to violate some fundamental concept: its designed to look more attractive to free software programmers, yet it encourages them to hope the company doesnt succeed.
www.bkbits.net is a free hosting service we provide.
It is not the same as openlogging, those logs are on www.openlogging.org.
I got an OK from Linus to put his trees on linux.bkbits.net,
you may go poke at them there.
Note that bkbits.net sort of looks like it might evolve into sourceforge
but that's not our intent. What we want is to provide that infrastructure
so that different people can host their own projects.
bkbits.net is a cache, you go there to fill your cache but then you have everything.
BitKeeper replicates the metadata so you are nowhere near as reliant on a centralized server
like sourceforge.
Since I was almost there, I'll pick up where I left off (and where you agree): namely, people have to be compensated for at least most of the time they spend writing their software.
The catch is that in order to demand compensation, you have to be able to withhold supply. There are two ways of doing this: the product method, which is that you pay me or you get no copy of the software, or the service method, which is that you pay me or I stop working on the product.
The tricky part is that once someone has the product in source form, chances are they care a lot less whether you can continue working on it or not. So giving away your source weakens your ability to demand compensation, and in effect reduces you to asking politely.
Don't get me wrong, I understand that the concept of selling Free software is relatively new, and models are being explored, and some degree of success seems to be occurring amid the many simultaneous failures. I really hope that a solid model is hit upon and proven capable of supporting software development at the same level as the closed-source models. I firmly believe that one should "free" one's code whenever possible, and successful business models built on Free software will make sharing possible more often. Everybody likes to share. (Well, not everybody, but I do.)
My objection to the post to which I originally replied lies with the idea that keeping your software closed is immoral "hoarding". Hoarding is taking a resource that was at some point available to all and collecting it and keeping it all to yourself. It's a valid view of software patents IMO (anyone could have had your idea, and still might, independently), but not of closed source in general (where there's a product of time spent laboring). It's also extremely arrogant and pretentious to descend from on high, survey the scene, and proclaim that the practice of extending an economic model developed for trade of physical items to the more ethereal world of software is "immoral". It's an outgrowth of the way trade has been conducted since the beginning, a natural extension into a new domain. It's also what's worked and continues to work, and is often very fair to all those involved.
I do get it, fully. I just find the rhetoric of "good" Free Software versus "immoral" closed software a bit short-sighted and fairly obnoxious.
quote from conclusion that sums up the essay (and told me everything I needed to know about the license/product):
Sometimes it is tempting to sacrifice our rights and freedoms for convinience, but we should not do so. There are many problems with CVS and other Free source management packages, and it would be nice to move to a more robust and more well-designed tool. We are better off to repair or fix the tools which are free, or if that is not acceptable to create new free tools that preserve the the rights and freedoms we enjoy.
I encourage BitMover to adopt the minimum requirements for freedom or openness that this community has defined, but at the same time I respect their wish to preserve the business model of their choice. At the very least I would like to see the rights the BitKeeper license does grant preserved and not subject to arbitrary revocation. I do hope that they will find some way to provide the community a truly free version of their tools and still meet their business goals.
I also encourage Free Software hackers to not use or stop using BitKeeper in their own projects. It might not be as convinient to use other tools, but in the long term we should be more concerned with preserving those rights and freedoms we currently exercise and enjoy daily. I personally have stopped using BitKeeper for all projects and have moved these projects back into CVS repositories. I hope that if you or your team is considering using or currently using BitKeeper that you will think about the implications of doing so and reconsider.
I encourage the entire community to support the efforts of Free and Open Source projects in this area. Source management is complex, and the efforts of the community to support Free and Open Source projects (CVS and Subversion are two such projects) are the best way we have to improve our development infrastructure.
source: Jack Moffitt's "A Critique of the BitKeeper License"
Use my userscript to add story images to Slashdot. There's no going back.