Interview with Andrew Tanenbaum (LinuxFr.org) [LWN.net]

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 17, 2011 22:44 UTC (Thu) by jgd (guest, #19925)
In reply to: Interview with Andrew Tanenbaum (LinuxFr.org) by cmccabe
Parent article: Interview with Andrew Tanenbaum (LinuxFr.org)

Tanenbaum gets small things right, big things wrong IMNSHO

BSD License problems: If you weren't there at the time you have no idea how bad it was. AT&T won a court case which not only prohibited the distribution of BSD but compromised the employability of every UC Berkeley CS graduate because they had been mentally contaminated by their exposure to AT&T licensed code. It took years of time and millions of dollars of public money to free BSD and to reverse that case.
Developer community, Copyleft. Open Source Free Software: Tanenbaum completely doesn't get the concept of community-based development. Linus didn't write much of Linux, he coordinates the work of tens of thousands of developers worldwide. There's nothing else like it. Tanenbaum doesn't get Open Source since he thinks it's great to have our lives (cars, appliances, etc.) managed by systems based on Minix but doing who-knows-what behind the barrier of closed source and DMCA. Tanenbaum does not seem to get the key difference between Open Source vs. Free Software and how the latter stimulates participation and protects the rights and security of users. The GPL License of the Linux kernel easily outweighs any possible technical superiority of any DMCA-protected proprietary kernel.
Upgrading kernel without rebooting: "Certainly Windows and Linux can't do this." Unless you use KSplice!
On being ahead of your time: It's nice that he "published a paper in 1978 on "something very close to the Java Virtual Machine". Maybe he "never got much credit for it" because the UCSD P-System was already doing it. The UCSD P-System was rapidly moving to dominate the world when it got saddled with a proprietary closed-source license and all development stopped - just the fate that almost happened to BSD.
Formal Verification: Tanenbaum dismisses the importance of the formal verification that has been accomplished by other projects for some microkernels and O/S components. Of course formal verification does not tell you when hardware interactions will bring you down. All it does is force the developers to keep the code squeaky clean and help them get rid of nearly all software bugs. Now that can't be very important, can it?
Missing what's truly important: Someone once asked Dennis Ritchie what was the best way to write a large C program. His answer: "Don't!". I am very much in favor of constructing ambitious software systems out of manageable and verifiable components interacting through simple, well-designed communications protocols. Minix has been and continues to be an important project in this area. The reason Minix has not been more successful up to this point are (1) initially getting off on the wrong foot with a proprietary license (2) the current lack of invitation and welcome of the participation of others and (3) the appearance (at least in this interview) of the lack of appreciation for the technical and other innovations of others.

(Log in to post comments)

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 0:29 UTC (Fri) by drag (subscriber, #31333) [Link]

Copyright itself is a problem. BSD underestimated it or did not consider the implications that the USA government could just one day decide to take a gigantic shit on their parade and ruin decades of hard work 'just because'.

This is why people that say IP rights is necessary for advancement in technology need to be smacked about the head and shoulders with the cluebat.

The copyleft license is a good work around for the evils of copyright. I think the effects of community building and corporations knowing they can release code certainly exists, but I think it tends to be overrated and the legal overhead and difficulties with copyright (even when largely neutered with copyleft) kinda makes it a very mixed blessing.

And, frankly, he is absolutely right that if it was not for the BSD lawsuit then we would all be cheerfully running BSD. And much better off for it, also. Instead of having to spend years developing Linux to get it to the point were BSD was before the lawsuit the Linux hackers would of instead be already working on something very capable and we would be far and advanced then were we would be now.

It wouldn't be the BSD we have today, of course. Quite possibly it wouldn't be much different then what we will have in a few years with Linux.

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 10:31 UTC (Fri) by kragilkragil2 (guest, #76172) [Link]

BS. :D

The BSDs we would have wouldn't be nearly as good as the BSDs we now have. There would have been hundreds of forks that never share their code and the core BSDs would have mostly stayed the same. Only the competition with fast evolving Linux made the BSDs progress like they did.

All IMNSHO of course. Without a timemachine it is hard to verify and you can argue about it all day. I just wanted to say that Linux is way ahead of everything the BSDs would have given us and that your argument is BS.

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 12:52 UTC (Fri) by trasz (guest, #45786) [Link]

Just like we have hundreds of forks of PostgreSQL, X.org, SQLite, Vorbis or any other BSD or MIT-licensed software, right?

Seriously - the "copyleft prevents forks" is one of the most stupid Open Source myths out there.

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 15:19 UTC (Fri) by pboddie (subscriber, #50784) [Link]

Just like we have hundreds of forks of PostgreSQL, X.org, SQLite, Vorbis or any other BSD or MIT-licensed software, right?

How would you actually know? That's the point you're missing here.

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 17:31 UTC (Fri) by bronson (subscriber, #4806) [Link]

You can't know. Why would you want to?

Should Richard Hipp care about all the sqlite forks embedded in proprietary devices all over the planet? They're not hurting anything.

(note for license anglers: I am not saying that MIT is better than GPL or anything like that. Both have their uses. I'm just wondering why people would want to follow every closed or inconsequential fork of their code)

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 23:52 UTC (Fri) by pboddie (subscriber, #50784) [Link]

You can't know. Why would you want to?

You can't know. That's the point. Why would you want to? To put your finger on the amount of effort going into forks that won't benefit the community.

Of course, people can fork copyleft-licensed code - it's the privilege everyone who gets the code has, after all - and in the event of that code reaching the community, it's possible that the community just decides that it isn't worth incorporating, but at least that avenue of potential incorporation of the work exists.

In general, the incentive to maintain forks is dependent on an organisation's size and willingness to collaborate. If you're a small organisation who doesn't want to share your code, you might favour permissive licences, but you're going to be kept busy maintaining your private changes on top of the community's work. Thus, enlightened organisations try and share changes even with permissively-licensed projects that don't insist on such sharing.

In contrast, large organisations who don't want to share can simply pick up a project and outrun the community by throwing developers at the code. When the licence permits the ability to maintain a private fork, a significant fork of the original project is thereby established. And that avenue of potential incorporation is removed, so there is nothing the wider community can do about the situation.

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 19, 2011 1:48 UTC (Sat) by elanthis (guest, #6227) [Link]

It takes no effort to use and fail to contribute back to a GPL project. First, nothing says you have to publicly announce the availability of the code. So long as the legalese nobody reads points out the GPL and you make some password protected FTP server available upon request, you comply. Second, nothing says your changes must be palatable to anyone outside your company. For an example, see Android or just about any other Linux driver or fork and all the crap they're filled with that no other developer wants to deal with. Finally, nothing says that the code must be easy for others to grab. You can just put up tar balls of code releases with no SCM history and be good, so long as there's comments or a ChangeLog file to comply with the GPL's (oft ignored) clearly marked changes clause.

A company using permissively licensed code can either be a good community member or it can be a jerk. A company using GPL licensed code can either be a good community member or it can be a jerk. Hell, a company using non-FOSS software can still be a good community member to the best of its ability (based on the specific terms of the software's license), or it can be a jerk.

If you want to argue that the GPL is important because it guarantees you can get the code to share free copies or make critical fixes or just research implementation, that's cool and makes plenty of sense. If you want to argue that the GPL in any way forces companies to meaningfully contribute back, you haven't paid even the slightest bit of attention to any major GPL-licensed project.

GPL is important because it keeps people honest...

Posted Nov 19, 2011 9:45 UTC (Sat) by khim (subscriber, #9252) [Link]

Companies fork the GPL all the time for various reasons. But if something in their fork is interesting and popular then someone else can bring this piece upstream. And this happens quite often (a lot of Linux drivers were created this way: initial awful driver by vendor was redone by someone else).

BSD license gives you no such advantage and GPL license with CLA does not give you such advantage. Ironically enough highly praised dual licensing model works as poisonous pill here. It's not zero-sum game because copies sold under proprietary license at least pay for a development. The worst kind is GPL with CLA and no dual-licensing. That's why even FSF only does this sparingly...

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 19, 2011 17:03 UTC (Sat) by pboddie (subscriber, #50784) [Link]

If you want to argue that the GPL in any way forces companies to meaningfully contribute back, you haven't paid even the slightest bit of attention to any major GPL-licensed project.

I didn't argue that it "forces" anyone to do anything. That's why I wrote that an "avenue of potential incorporation" of modifications to a project exists, as opposed to a mechanism to compel organisations to contribute changes upstream.

And much of your response actually restates what I wrote, in fact, which I suppose I should take as some kind of compliment.

Interview with Andrew Tanenbaum (LinuxFr.org)

Posted Nov 18, 2011 15:28 UTC (Fri) by andresfreund (subscriber, #69562) [Link]

> Just like we have hundreds of forks of PostgreSQL, X.org, SQLite, Vorbis or any other BSD or MIT-licensed software, right?
You do realize that there are - and were - several fork of postgres around? Enterprisedb, Greenplum, Aster, ....