[TUHS] Free/NetBSD revision history (was Code bloat)

[imp@bsdimp.com (Warner Losh)]

On Thu, Feb 9, 2017 at 4:38 PM, Jacob Goense <dugo at xs4all.nl> wrote:
> On 2017-02-09 11:14, Warner Losh wrote:
>>
>> I thought someone had posted a github project to merge the history of
>> all publicly available sources of unix.
>
>
> That's the thing, it banks on what is publicly available.
>
> NetBSD and FreeBSD both started out from 386BSD + patchkits. They threw
> it is cvs, and engineered a release.
>
> This was all, in essence, Net/2 based, then the USL vs. BSDi lawsuit
> kicked in and was settled by a.o. "encumbering" Net/2 by agreement.
>
> Panic ensued and the NetBSD and FreeBSD teams took a chainsaw against
> what they had released until then.

That was actually part of the agreement with AT&T to end the hostilities.

NetBSD did it by butchery. FreeBSD did it by reimporting from 4.4lite,
basically (the basically part is a bit messy).

> The publicly available repos from that period are butchered.

True. I had thought the original FreeBSD 1 repo was now publicly
available. I know I can get copies of it as a FreeBSD project member.

> The number of people on earth trying to curate stuff like the history
> of locore.s/tty.c between 386BSD and the reboots of Net/FreeBSD is a
> handfull, and I'm being optimistic here.

Yea, the FreeBSD CVS tree I think has that history. It started out
life trying to make the patch-kit to 386BSD make sense as dealing with
a boatload of patches soon grew unwieldy as the number proliferated
and you started getting patches on patches.

> This stuff has been deliberately purged and hard to find. Jason Stevens
> went as far as reconstructing a NetBSD 0.8 kernel because the complete
> sources where nowhere to be found. Then he ran into the proverbial
> coughing, chain smoking guy in a raincoat in a parking garage with
> a manilla folder of a CD-ROM of
> ftp://agate.berkeley.edu/pub/NetBSD/NetBSD-0.8,
> or was it a forgotten ftp site?

Don't know anything about that... But the Truth is Out There.

> Anyway, the revision history of the "encumbered" pieces in NetBSD is
> probably lost, but at least the 0.8 checkpoint was unearthed.

I thought it was just shielded from public view and many of the NetBSD
folks had copies. Could be wrong though.

> If you take a close look at the publicly available revision history of
> FreeBSD you'll notice some serious gaps as well. Someone went through
> that cvs with an axe or surgical knife for legal reasons (and made a mess
> teleporting AMD64 to the early 90s).

Yea, CVS doesn't support repo-copying for crap. But it was done to go
from i386 to amd64.

> What dspinellis did with git is truly awesome. But I see the scars the
> USL vs. BSDi lawsuit made. I have no idea why I care, but I do. I respect
> that not everything can be made publicly available, but I pray stuff
> such as an original FreeBSD revision history is at least dumped into
> hidden archives like Warren and friends keep until the time is right.

The old CTM archives might have stuff, it that was up and running
before the lawsuit.

I know that the FreeBSD 1 archive exists in multiple places.
Compressed it is 18MB, or 185MB uncompressed (clang's history is
bigger than that, though checked out it is only about 131MB).

Warner