Hacker News Reader

Comment by hyperpape | Hacker News Reader

I went back and looked at the older discussion, and it doesn't paint Stallman very well as the head of a project. He pins the question of whether to keep using bzr not on whether it is good or whether the Emacs developers want it, but on whether it's being "maintained".

But then he seems to define maintenance as having fixed a specific bug that's been around for over a year, blocking a point release.

He admits that he can't follow the developers list to see if they're genuinely doing active maintenance (reasonable enough: he has a lot on his plate), but also won't accept the testimony of Emacs developers that the mailing list is dead and there's no evidence of real maintenance.

When questioned, he says that there's too much at stake to abandon bzr if it can be avoided at all. But the proposed replacement is GPL software. This is just madness.

Refs: http://lists.gnu.org/archive/html/emacs-devel/2013-03/msg009....

http://lists.gnu.org/archive/html/emacs-devel/2013-03/msg008...

(and surrounding posts).

View on HN

I think this is more inertia than anything else. It is pretty obvious that Stallman isn't the easiest person to talk to, but he has consistently put huge amount of efforts into GNU for lots of years, and, whatever he did as the maintainer, seems to have worked out pretty much in the end. If there was anyone as active in the Emacs community today as Stallman was in his prime, they could easily push through changes despite Stallman's disagreement, via a fork or otherwise, but there isn't such a person, and that is that.

Consider that it has been almost 30 years since GNU Emacs started, for most of this time RMS has been the maintainer, and GNU Emacs continues to this day to be the most advanced, popular and active Emacs there is, while the various forks in existence like XEmacs, SXEmacs lost steam pretty quickly. So it certainly hasn't been all bad on his side.

The popularity of a specific fork of Emacs is a moot point, it's the popularity of Emacs as an editor (or should I say, as a platform?) that is at issue. Stallman was active, how that he's not active he needs to either defer to people who are active or watch as fewer and fewer people take advantage of his work.

Stallman doesn't have much influence on Emacs development itself anymore, but Emacs is a GNU project, Stallman is still the head of GNU, and a move to another VCS would be a major organizational change. I too wish he would be less hard-headed about this, but lets not overblow this issue out of proportion.

In replies, he says he doesn't oppose the move to git. Most devs who've replied support git. There's already a mirror. Let's not make mountains out of molehills...

Over those 30 years the emacs codebase must have moved through several revision control systems. Why is moving to another one such a big deal?

Because it's important to have a cohesive set of packages that are parts of the GNU project. It's about culture and cohesion, not just licensing.

As he said, 'more than Emacs is at stake here.'.[1]

I presume this refers to Bazaar's status as part of the GNU project, and that RMS did not want to write off part of GNU without being certain he needed to.

Regardless, he has since OKd the switch from bzr:

I don't insist that Emacs should stay with bzr. I chose to support bzr because it was still a contender at the time. [2]

[1] https://lists.gnu.org/archive/html/emacs-devel/2013-03/msg00... [2] https://lists.gnu.org/archive/html/emacs-devel/2014-01/msg00...

edit: formatting

I think Stallman is being reasonable here.

He's not always an unreasonable man.

https://paragraft.wordpress.com/2008/05/01/progress-and-the-...

I think the subsequent OK'ing is unfortunate. Attracting younger hackers shouldn't be a more important goal than integrity.

What does this have to do with integrity?

I mean, don't not fixing what ain't broke is certainly a matter of integrity. Fixing something to the latest hippest new fad just to "attract young hackers" detracts from integrity. Making the editor an awesome badass attractive editor should be enough. Otherwise you'll only be attracting groupthink-prone douchebags, anyway.

For better or worse, having a vision is a very different skillset from having the management skillset to make things happen. Sometimes people deride management when talking about the importance of vision and leadership. There is also the current trend of companies giving up managers entirely.

RMS is a prime example of why management is an important skill. It is inefficient for him to do everything himself. It's also inefficient for him to chase people down for all the details. If he had a solid COO who understood his vision, his organization would be much more effective.

> it doesn't paint Stallman very well as the head of a project

Is this news to you?

Cf. e.g. http://www.jwz.org/doc/lemacs.html

shrug More confirmation.

Not to mention his insistence on GNU Hurd being based on Mach, which pretty much ended up killing the project.

That's a simplistic view. First, the GNU project is very much alive: the GNU tools are used in a huge number of operating systems and are installed on a staggering number of devices. I would bet that the system you are writing this comment from is running thanks to the GNU software.

Second, it is debatable whether sticking to Hurd was a good or a bad idea technically. Imagine if Stallman and co. managed to convince a good number of developers that it was a good idea and the kernel was competitive with Linux, BSD's, etc. If you believe you have a technically superior vision for your product, should you compromise on it just because people who do not share in your vision will not join you?

In the end, I think what killed the pure GNU/Hurd OS was bad PR and absolutism. Hurd as a technical question was just a small part of that. Remember, the debate between the Free and the Open Source guys was pretty fierce. Today we use terms like FOSS to describe all open software, but when Linux and Hurd were young these were different camps with opposing philosophies, and the one that appealed to more developers won out. In simplistic terms, you can think of this as the VHS vs Betamax debate. Can you blame the Betamax backers for continuing to try to push it and "killing" it as a result?

Wait. You misunderstood me. I didn't say the entire GNU Project is dead. Hell no. When I said "project", I was referring specifically to GNU Hurd.

Also my statement was going by the words of the Hurd's former project leader Thomas Bushnell:

"RMS was a very strong believer -- wrongly, I think -- in a very greedy-algorithm approach to code reuse issues. My first choice was to take the BSD 4.4-Lite release and make a kernel. I knew the code, I knew how to do it. It is now perfectly obvious to me that this would have succeeded splendidly and the world would be a very different place today.

RMS wanted to work together with people from Berkeley on such an effort. Some of them were interested, but some seem to have been deliberately dragging their feet: and the reason now seems to be that they had the goal of spinning off BSDI. A GNU based on 4.4-Lite would undercut BSDI.

So RMS said to himself, "Mach is a working kernel, 4.4-Lite is only partial, we will go with Mach." It was a decision which I strongly opposed. But ultimately it was not my decision to make, and I made the best go I could at working with Mach and doing something new from that standpoint.

This was all way before Linux; we're talking 1991 or so." [1]

http://www.groklaw.net/article.php?story=20050727225542530 [1]

Note, in regard to Bushnell's quote, that "1991 or so" and "way before Linux" are contradictory. (Or, at least, require an strained definition of "way before"; Linux was first released in 1991.)

>the GNU tools are used in a huge number of operating systems

You mean linux? That isn't a huge number.

>and are installed on a staggering number of devices

The staggering number of devices you refer to almost exclusively run busybox or one of the similar projects. GNU software is hugely bloated and not a good choice for embedded systems.

>Second, it is debatable whether sticking to Hurd was a good or a bad idea technically

No, it was fine technically. It had no developers and so nothing happened. Minix exists, obviously microkernels are possible.

Last I heard a lot of users of Solaris and other Unixes often also use the GNU tools, though those are generally dying out in favour of Linux anyway.

Last you heard wrong then. Occasionally we begrudgingly install some GNU bloatware because some poorly written software requires it. That's about it. Every other OS already comes with its own versions of all the unix tools.

GNU tools were often used on other OSes too including Solaris and other commercial Unixes but also Windows under Cygwin. Particularly GCC, make etc. but also command line tools such as grep where the default platform versions were often not as feature rich. I didn't use those platforms so others will remember and know better but I don't think huge was an obviously wrong description.

There is a big difference between "some people optionally could install GNU stuff in addition to their existing tools" and "those OSes use GNU tools".

>the GNU tools are used in a huge number of operating systems

Fair enough, I read "in" as a synonym for "on" but read in a less casual way the difference could be important.

Even in the "in" case I think many BSD's used GCC as the default compiler (CLANG seems to be taking over now).

Apple still insists on building their kernel atop Mach, and it doesn't seem to stop their momentum. (Even though it's kind of a strange choice.)

From what I understand, the Mach kernel which is now used in XNU is not the Mach micro kernel (3.0 >) but based upon the pre-micro-kernel 2.5 version of Mach.

I'm not sure where I read this originally but I just googled this source which seems to back it up:

http://www.roughlydrafted.com/0506.mk3.html

I am not intimately familiar with the history of Mach, but what I do see is that in the late 80s and early 90s these two groups (NeXT and GNU) seeing Mach as the future (a position that makes no sense at a later time), and having vastly different outcomes.

I don't know much about how these people work, but I always figured Mach at Apple is just about momentum and familiarity of contributors, rather than technology. NeXT hired Tevanian who worked on Mach at CMU, they spent roughly a decade hacking on Mach, then Apple did the same. I'd imagine they employ people who know Mach well and haven't seen it as worthwhile to replace it.

I even remember they had this goofy project "MkLinux", which sought to put Linux in the position that BSD carries with XNU, on top of Mach... Just goofy stuff, unless you figure they had Mach hackers on staff.

According to the Apple docs XNU is based on Mach 3.0.

https://developer.apple.com/library/mac/documentation/Darwin...

That's misleading. Apple merged in a bunch of Mach 3 code but still maintains the architecture of Mach 2.5 (i.e., Mach + BSD both running in supervisor mode in one big monolith; no BSD server; xnu is not a microkernel…).

It is odd, I predicted about 5 years back that Apple would gradually converge the API to FreeBSD and then switch. They would still be better off doing that I think, but they show no signs of doing this.

Why would they do that? Mach allows for some cool stuff, like kernel extensions being isolated.

Because building and maintaining a competitive OS is a huge effort, and FreeBSD has lots of cool stuff which they can't use...

It's a huge effort worth that seems to be paying off quite handily. XNU seems like a competitive edge over the monolith FreeBSD, as far as desktop and mobile OS is concerned. FreeBSD has plenty of cool stuff, but in a completely different domain.

> XNU seems like a competitive edge over the monolith FreeBSD, as far as desktop and mobile OS is concerned.

This seems a bit delusional. I don't think it's controversial to say that Apple's biggest differentiators exist at higher levels than kernel space. I'd go so far as to say that anyone who claims that Apple's success is rooted in XNU and that the same could not have been done with Linux or *BSD at the lowest layer and all other pieces being equal does not understand what a kernel is.

I don't know anyone saying Apple's success is rooted in any one thing. However, IOKit is very important, "common code for device drivers, this framework also provides power management, driver stacking, automatic configuration, and dynamic loading of drivers". Even if the entire OSX could have been implemented on top of Linux and FreeBSD reusing those kernels (and it does reuse FreeBSD for low level POSIX apis!), how productive would that have been? I honestly don't know, but I choose to trust what I read from the original developers.

I was assuming they would port iokit as part of it...

Those other kernels have driver frameworks too. And if they find them inadequate for some reason, they are of course able to make modifications.

But this is not a suggestion for them to scrap it, necessarily. As I said in some other comments on this thread, I think the real reason is that they had people that knew their existing kernel well, and don't see a need to replace it.

FreeBSD does not have Mach ports (the IPC system). There is a ton of what OSX does that needs this.

See my comment up thread about Jordan Hubbard's plans.

Needs mach ports, meaning can't be simulated with Unix domain sockets or something else? I doubt it.

> XNU seems like a competitive edge over the monolith FreeBSD,

xnu is also a monolithic kernel, just with some nice message-passing primitives. Think Mach 2.5.

Darwin + Aqua also has lots of cool stuff which *BSD can't use. I can see a competitive advantage in making sure that remains the case.

xnu kexts are not isolated and probably never will be, although IOKit exposes some things to userspace.

They're isolated enough that drivers can crash and just be reloaded, without a kernel panic.

No, they are not. A crash in a kext causes a kernel panic.

You are right. An actual crash in kext causes kernel panic. I think I read that this is was possible, but I can't find how, and when, right now. The common thing that I'm thinking of is just voluntary error (exception) handling in kexts and reloading of kext on their own. There is some isolation and benefit to it.

Edit: ok, I just figured out my source of confusion. IOKit allows userspace drivers, which can crash without resulting in panic.

IOKit allows communication between kexts and userland. You could call that "userland drivers", but it's not like you can write only userland code to implement a driver.

Amazingly, jkh is talking about bringing More parts of Mach to FreeBSD. To make it more like OSX.

Back when the choice was made, micro-kernels was all the rage in both academia and commercial ventures, and Stallman chose Mach since he thought it would speed up development, he was hardly alone in choosing Mach at this time, Apple (MkLinux, NeXTSTEP), IBM (Workplace OS) amongst others.

He fully acknowledged that he made a mistake in going with Mach and as soon as Linux took off FSF focused on providing the necessary software to combine with Linux into an operating system and placed Hurd on 'life support', where it's been ever since.

IIRC, ARPA at some point was more interested in funding projects based on Mach than based on any other kernel.

I'm not sure that conversation is a good example of what you are trying to describe.

Let's see if someone did the following things to the python project:

1#: Hire away the package maintainer. Then rather than continue and finish any current work, effectively remove that person from the community project.

2#: Redesign underlying structure of the project (like say, PyPy), but don't discuss any changes with the community. No PEPs, no discussion on mailing lists, no communication what so ever.

3#: Ignore current list of new feature being worked at. Community goals are unimportant.

4#: Add code regression! Do not care about maintaining performance.

5#: Demand that the changes get implemented immediately in next official release.

Would anyone expect that to actually work today? Sure, Stallman could be more diplomatic and find (and succeed) with a middle ground solution, but the above steps are not how you join an ongoing software project.

I tried to read that conversation in a neutral light, because I think RMS and JWZ are both kind of ... polarizing personalities, but RMS really came off as a stubborn, ineffective whiner in that whole thread. In particular "you hired away my maintainer" is a pathetic excuse for not releasing something for so long. Anybody could have hired away your maintainer, and you'd have to soldier on; it has no relevance that your maintainer went to a "competing" (in your territorial view) project.

I also read the argument about the redesign of the event system and was pretty flabbergasted. The argument seems to reduce to "lucid emacs decided to design a proper event datatype because having an event be entirely represented by a simple integer keycode both lost information and made it impossible to represent certain keystrokes" versus "but ints are simple and backward compatible!"

Funny, I had the opposite view. I guess your opinion of the personalities really does change everything.

There is one (and only one) other possibility, which is that you and I both read the conversation with flawless objectivity, and that you are wrong. :D

Poaching the lead dev to work on your own fork is a hostile move, and in my book, it provides a reasonable excuse for the delays, since nobody was able to take his succession immediately.

The guys at Lucid also barely communicated for long periods of time, making collaboration impossible.

> Poaching the lead dev to work on your own fork is a hostile move

That's an argument one can use when asked to spend more time with kids.

> it provides a reasonable excuse for the delays, since nobody was able to take his succession immediately

Huh?!? Open Source model not working or what?

>> it provides a reasonable excuse for the delays, since nobody was able to take his succession immediately

> Huh?!? Open Source model not working or what?

I guess its only illusion and crazy lawyers who like to add anti-poaching clauses in employee contracts. If a company can't handle loosing their top engineers, clearly its the the proprietary model that is not working.

> I couldn't change the plans, so I had to make the best of them. I suggested a design to him, one oriented toward editing formatted text--a feature I wanted Emacs to have, eventually.

Hah, 20 years in the making!

the exchange made lucid seem like a bunch of assholes

Care to explain? I thought that the Lucid guys made reasonable technical arguments (especially in light of, y'know, history, over the last twenty years) and that RMS was attempting to both grandstand and emotionally manipulate people into adopting his preferred position.

I thought RMS had resigned as Emacs’ maintainer?

http://lists.gnu.org/archive/html/emacs-devel/2008-02/msg021...

Maybe--but the tenor of the discussion strongly suggests, if not implies, that he makes the call about whether they can leave bzr.

One of the reasons I chose GNU Emacs over XEmacs was a feeling that GNU Emacs will be maintained, even advanced, for as long as RMS has the strength to type. It's his baby.

(There were other reasons, the big one being momentum. XEmacs didn't run on the platform I was on for a long time, so it would have been a switch.)

An excellent point; for example, although I don't have any reason to think RMS himself did this, Emacs 24.4 will have file change notification support across all platforms where it compiles and where such notifications are available.

I grant that's a somewhat overdue feature for Emacs to have (e.g., I've wanted it ever since I set up Dropbox to synchronize my org-mode files across all my boxes), but it's definitely evidence that Emacs isn't lacking of maintenance and improvement.

I wonder: If Emacs rarely gains new features in core, is it because there aren't enough developers doing enough to improve it, or rather because people are having a hard time thinking up new features to add which Emacs doesn't already have?

Lisp systems legacy I suppose, the core is ~tiny, everything else is a library.

Sorry, I was unclear; by "core" I mean "the Emacs distribution itself", as opposed to libraries you find on Github, EmacsWiki, or wherever else that isn't part of the standard Lisp library you get when you download the Emacs tarball.

RMS has RSI, so he doesn't type, he dictates. Oh, shit -- that didn't come out the way I meant it to. ;(

I wonder what software package he uses.

    "When questioned, he says that there's too much at stake 
    to abandon bzr if it can be avoided at all."

The big difference between then and now is that this time you have a very competent developer with an exceptional reputation offering to lead the migration. With esr leading things, there is a lot less at risk.

Reply on news.ycombinator.com