rob at landley.net
Sat Oct 8 12:04:15 PDT 2011
On 10/08/2011 12:55 AM, John Spencer wrote:
>> Oh I'm mirroring all the gplv2 stuff (and possibly some bsd stuff) at
>> I've spoken to Eben Moglen (co-author of both GPLv2 and GPLv3) about
>> this when I visited the SFLC's office in New York (which was handling
>> the busybox license enforcement lawsuits), and he _almost_ convinced me
>> that GPLv3 wasn't worth opposing.
> indeed, seems to much effort, or leaves you behind with outdated sw...
You just neatly summarized using Linux instead of windows on my laptop.
And before that, OS/2...
>> Computer history is a hobby of mine, and I wrote up some of the REAL
>> history on this topic (as I understand it) last year:
> interesting read.
> afaik unix systems were relatively uncommon here in europe and only got
> popular with linux.
> so you don't really learn about that kind of history.
The Ken Olsen interview is great (might want to read the PDP-8 FAQ
sections "What is a PDP" and "What is a PDP 8" first).
Although if it's video you want, start with:
Based on Robert Cringely's book "accidental empires", which is great.
(Doesn't cover unix, but it gives great context for the restof the
industry". Steven Levy's book "Hackers" gives another important
sub-plot, and Ken Olsen's interview is sort of "the other side of the
story" of that book...)
Don't get me started about computer history. It is a FASCINATING topic...
>> That said, the third parties like Code Sourcery who actually write new
>> code push that code upstream under the project's current license, and
>> since the FSF successfully recaptured EGCS stuff like armv7 and
>> microblaze support is either in external forks or only in newer versions
>> of gcc/binutils after the gplv3 switch. And pcc and llvm/clang have a
>> long way to go to build a workable Linux system even under x86...
> oh, i actually expected clang to be mature enough to build the kernel.
The kernel built _only_ with gcc for the longest time, and developed a
lot of weird extensions and corner cases and dependencies on specific
inlining behavior. They cleaned it up a bit, but so far only four
toolchains I know of have built the kernel:
1) tccboot built a modified subset of 2.4 many years ago, but it was
never reproduced and the project is essentially no longer maintained.
2) Intel's ICC built it (by implementing the necessary gcc extensions),
but that's a proprietary x86-only compiler.
3) llvm/clang managed to build a fiddly subset of 2.6 (similar to
tccboot) one year ago:
So LLVM is making lots of progress (due to Apple's sponsorship).
I talked to the pcc guys about this a while back, google found this:
But there was a thread on the pcc mailing list that went into more
detail, i thought. Plus I linked to the recording of the OLS 2008 BOF I
didn't _MEAN_ to spend "hosting" but they kept asking me questions and
the camera _followed_ me when I wandered away from the front of the
room... *shrug* Covered a lot of this material there, it was an hour.
>> The FSF was threatening to sue this guy, which would have bankrupted him
>> long before a verdict.
> well, maybe, but they didn't sue him in the end.
They'd already told him they would if he didn't fold.
To quote Monty Python: "It's a nice army base you have here, colonel.
Shame if something were to happen to it." "All those tanks. Very
They don't actually have to set fire to a tank in order to be doing a
shakedown. That's like saying "the guy who raped that woman at gunpoint
didn't actually pull the trigger, therefore it was consensual"...
>> Trust me, the FSF was talking out of its ass here. It led to me washing
>> my hands of the SFLC:
That's why I stopped doing busybox license enforcement lawsuits, but
there was also the fact that from an engineering perspective it was
useless: we never got any code to add to the project. There wasn't any
worth bothering with, ever.
>> They did it because once again, they were utterly irrelevant and felt
>> left out, so they struck out blindly to grab the spotlight. The end
>> result was that CISCO DISSOLVED THE LINKSYS DIVISION AND GOT OUT OF
>> LINUX DEVELOPMENT ENTIRELY. (Outsourced what they couldn't immediately
>> kill to Red Hat, if I recall.)
> ok. thats somewhat unfortunate, if it really was only about the realtek
Erik Andersen's father is a lawyer. He and harald welte and some other
guys got together and browbeat Linksys' modified busybox and kernel
source code out of them back in 2003:
That wasn't an official public lawsuit because A) nobody was quite sure
it would win in court, B) they're expensive and protracted. Instead
this was working behind the scenes to put pressure on Linksys, but it
was "here's wha we could sue with".
Yes, it's a bit like with Mepis, the fangs were bared without actually
biting anybody. But they got a big payoff: open source embedded
routers. Linksys used Linux in the first place because Pauline
Middelink invented IP Masquerading on Linux (Cisco later renamed it
"Network Address Translation" but we had it first). What Linksys
brought to the table was cheap tiny hardware to run it on, in a simple
"drop it in and it works" package. We wanted to run our code on that
hardware, and to get their existing setup to modify we were willing to
sue. This resulted in OpenWRT and OpenEmbedded and so on. (When Erik
got the code he ported a uClibc-based debian to his router and ran
busybox.net and uclibc.org on it for a bit, and also did buildroot.)
But the downside of pressuring Linksys is that they stopped doing Linux
development for about five years (outsourcing what they had to Taiwan
and switching to some other embedded OS for most of their newer
routers). They were just taking it back in house around 2008 when the
FSF filed suit.
The FSF lawsuit turned "ok, we got sued into compliance and now we're in
the clear" into "we will never stop being sued over this piece of
technology". And there was NO UPSIDE because there was NO CODE THEY HAD
THAT ANYBODY ELSE CARED ABOUT. They _themselves_ didn't care about that
old toolchain anymore, they were only still using that ancient gcc 3.x
and linux 2.4 piece of crap because they hadn't been doing any
development in that period, and were moving to gcc 4.x and linux 2.6
with vanilla unmodified versions of everything. I was there helping
them make this transition, THAT was the project that got shut down by
the FSF's blundering.
Note that during this 5 year gap (between Linksys releasing source in
2003 and me visiting in 2008) the SFLC was formed, partly so Eben Moglen
could get real world experience for his law sudents at NYU, and partly
so he could distance himself from an increasingly insane and irrelevant
As always, there's a lot going on behind the scenes, which the FSF
either doesn't understand or is completely tone-deaf to, and they wind
up doing more harm than good. (If this _wasn't_ the case, Linux would
never have been needed, EGCS never would have happened,
cygnus/codesourcery would be unnecessary, the SFLC would still just be
the FSF's legal arm...)
The FSF takes credit for things other people do and screws up the stuff
they try to do in house. They're the reason Android has a policy "No
GPL in userspace". (Probably due to their GPLv3 blundering
specifically, but the policy pushes v2 out as well.)
I was actually quite the fan of GPLv2, before v3 came out:
More information about the Aboriginal