28 Apr
2025
28 Apr
'25
11:36 a.m.
http://xahlee.info/UnixResource_dir/writ/unix_origin_of_dot_filename.html says
I'm not sure but I believe .. went in during the
Version 2 rewrite
.. was there from the beginning. The v1 man page directory(v) says,
By convention, the first two entries in each directory
are for "." and "..".
Doug
29 Apr
29 Apr
1:21 a.m.
New subject: V7 ls and dot files?
So now I see that v1 had hierarchical path names. I thought that came in
v2. I stand corrected.
-rob
On Mon, Apr 28, 2025 at 9:36 PM Douglas McIlroy <
douglas.mcilroy(a)dartmouth.edu> wrote:
http://xahlee.info/UnixResource_dir/writ/unix_origin_of_dot_filename.html
says
I'm not sure but I believe .. went in during
the Version 2 rewrite
.. was there from the beginning. The v1 man page directory(v) says,
By convention, the first two entries in each
directory are for "." and
"..".
Doug
3:17 a.m.
New subject: V7 ls and dot files?
Aside from punched card stuff. I started life in TOPS-10. 78 vintage, not
very heavily modified locally as I recall. (people did run local mods, I
don't think we'd applied any)
I had forgotten TOPS-10 had a logical equivalent of a mount point, the MFD.
And I had forgotten inside the [xxx,yyy] group/user identity directory, you
could nest 5 sub-directories deep.
5! such humongous nesting! who could need more than 5 levels deep!
What I did remember is there was no evident ".." equivalent in the commands
to "be" in a directory. The idea of needing to go back to your parent,
positionally inside a subdirectory but without recourse to the actual path
down from the MFD appeared to be missing. I suspect whoever designed this
was living their best life in JCL and saw no interaction on paths as
needing "optimisations" = SET DIRECTORY obviously demanded a path down from
an MFD. Fool of a took!
By the time of VMS, you had the same $MFD:[path.path.path...] structure but
now up to 255 nesting levels deep, and the "-" nonce directory meant "the
parent to the one you are talking about" so at least in hypothesis a "cd
.." command was possible without needing to know the descent path. 255!
magic numbers! Who could need so many nesting levels?
When I got the great uplift to Unix in 82 and realized . and .. existed, It
was a very strong mind-officially-blown moment. Its like the zen koan for
the ages: Who the hell thought it was ok to design a filesystem ANY OTHER
WAY. I simply cannot handle that I was using some system (Norsk Data?)
around the same time and it didn't have cd .. -It still had a TOPS-10 class
view of the world as a descent tree only. Stone Age!
I still regret that the newcastle connections ... naming model didn't get
up. I think it was very nice.
G
11:28 a.m.
New subject: V7 ls and dot files?
I seem to remember that V7 was the first suystem which had a mkdir system call. That might
have changed what ls -a showed.
jaap
6:52 p.m.
New subject: V7 ls and dot files?
I thought so too, but found this reference to a Version 1 mkdir:
https://man.cat-v.org/unix-1st/1/mkdir
I recall reading a Version 6 or 7 man page about mkfs that included the
ability to populate a file system with some directories, and I thought that
implied that users couldn't create directories. The man page referenced
above hints that mkdir is run as the 'system user', presumably root, and
becomes the owner of new directories.
Rik
On Tue, Apr 29, 2025 at 4:28 AM Jaap Akkerhuis <jaapna(a)xs4all.nl> wrote:
I seem to remember that V7 was the first suystem which
had a mkdir system
call. That might have changed what ls -a showed.
jaap
7:18 p.m.
New subject: V7 ls and dot files?
On 2025 Apr 29, at 20:52, Rik Farrow
<rik(a)rikfarrow.com> wrote:
I thought so too, but found this reference to a Version 1 mkdir:
https://man.cat-v.org/unix-1st/1/mkdir
That is the command in man(1) and always existed. System calls documentation live in
man(2).
jaap
I recall reading a Version 6 or 7 man page about mkfs that included the ability to
populate a file system with some directories, and I thought that implied that users
couldn't create directories. The man page referenced above hints that mkdir is run as
the 'system user', presumably root, and becomes the owner of new directories.
Rik
On Tue, Apr 29, 2025 at 4:28 AM Jaap Akkerhuis <jaapna(a)xs4all.nl
<mailto:jaapna@xs4all.nl>> wrote:
> I seem to remember that V7 was the first suystem which had a mkdir system call. That
might have changed what ls -a showed.
>
> jaap
>
7:33 p.m.
New subject: V7 ls and dot files?
On Tuesday, April 29th, 2025 at 12:19 PM, Jaap Akkerhuis via TUHS <tuhs(a)tuhs.org>
wrote:
What I see in the history is:
V1 - mkdir(1) and mkdir(2) are both present, mkdir(1) indicates that:
On 2025 Apr 29, at 20:52, Rik Farrow
<rik(a)rikfarrow.com> wrote:
I thought so too, but found this reference to a Version 1 mkdir:
https://man.cat-v.org/unix-1st/1/mkdir
That is the command in man(1) and always existed. System calls documentation live in
man(2).
jaap
>
> I recall reading a Version 6 or 7 man page about mkfs that included the ability to
populate a file system with some directories, and I thought that implied that users
couldn't create directories. The man page referenced above hints that mkdir is run as
the 'system user', presumably root, and becomes the owner of new directories.
>
> Rik
>
>
> On Tue, Apr 29, 2025 at 4:28 AM Jaap Akkerhuis <jaapna(a)xs4all.nl> wrote:
>
> > I seem to remember that V7 was the first suystem which had a mkdir system call.
That might have changed what ls -a showed.
> >
> > jaap The standard entries "." and ".."
are made automatically.
and this text persists through to V6 at least.
Whereas mkdir(2) states:
The special entries "." and ".."
are not present.
With V2, mkdir(2) is renamed makdir(2) but still retains this text. In V4, this is
replaced with mknod(2) which drops this text. So for the purposes of . and .. the
situation should be unchanged from V1 to V3, with V4 switching to mknod(2). This is all
from my mandiff project which is paused at V6 until a few more manuals from the time
(namely USG PG-III and PWB/2.0) manage to crop up. Comparing between V6 and V7 should
lend more clarity to the situation.
- Matt G.
7:27 p.m.
New subject: V7 ls and dot files?
On Tue, Apr 29, 2025 at 3:01 PM Rik Farrow <rik(a)rikfarrow.com> wrote:
I thought so too, but found this reference to a
Version 1 mkdir:
https://man.cat-v.org/unix-1st/1/mkdir
I recall reading a Version 6 or 7 man page about mkfs that included the ability to
populate a file system with some directories, and I thought that implied that users
couldn't create directories. The man page referenced above hints that mkdir is run as
the 'system user', presumably root, and becomes the owner of new directories.
That's the man page for the `mkdir` command, but I think the
distinction was for the system call. Before the introduction of
`mkdir(2)`, directories were created by `mknod(2)`, and that was true
through the 7th Edition and into 32/v.
I would have thought that `mkdir` came with FFS and indeed, it's not
in 3BSD or 4.0BSD; looks like it came with 4.1c.1, which is where UFS
starts to be integrated.
- Dan C.
On Tue, Apr 29, 2025 at 4:28 AM Jaap Akkerhuis
<jaapna(a)xs4all.nl> wrote:
>
> I seem to remember that V7 was the first suystem which had a mkdir system call. That
might have changed what ls -a showed.
>
> jaap
>
30 Apr
30 Apr
7:09 a.m.
New subject: V7 ls and dot files?
Jaap Akkerhuis <jaapna(a)xs4all.nl> wrote:
I seem to remember that V7 was the first suystem which
had a mkdir system call. That might have changed what ls -a showed.
jaap
Thanks. IIRC it was 4.2 BSD. V8 may have picked it up.
Arnold
23
days inactive
25
days old
8 comments
8 participants
participants (8)
-
arnold＠skeeve.com -
Dan Cross -
Douglas McIlroy -
George Michaelson -
Jaap Akkerhuis -
Rik Farrow -
Rob Pike -
segaloco