From mboxrd@z Thu Jan  1 00:00:00 1970
From: dexen deVries <dexen.devries@gmail.com>
To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net>
Date: Thu,  3 Feb 2011 15:27:59 +0100
User-Agent: KMail/1.13.5 (Linux/2.6.37-rc6-18+; KDE/4.5.5; x86_64; ; )
References: <201102031245.33842.dexen.devries@gmail.com>
	<201102031444.40895.dexen.devries@gmail.com>
	<AANLkTimje_peQjPjpCU0bVcAL6SAH1k1P0NSUmGMZAU4@mail.gmail.com>
In-Reply-To: <AANLkTimje_peQjPjpCU0bVcAL6SAH1k1P0NSUmGMZAU4@mail.gmail.com>
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Message-Id: <201102031527.59587.dexen.devries@gmail.com>
Subject: Re: [9fans] files vs. directories
Topicbox-Message-UUID: aad3ce24-ead6-11e9-9d60-3106f5b1d025

On Thursday, February 03, 2011 03:15:29 pm roger peppe wrote:
> On 3 February 2011 13:44, dexen deVries <dexen.devries@gmail.com> wrote:
> > On Thursday, February 03, 2011 02:36:40 pm roger peppe wrote:
> >> On 3 February 2011 11:45, dexen deVries <dexen.devries@gmail.com> wrot=
e:
> >> > read(open("/foo")) returns byte stream under entry `foo' in the root
> >> > object.
> >> >=20
> >> > readdir("/foo") returns `bar' (and possibly others) -- entries in
> >> > hierarchical section of object `/foo'.
> >>=20
> >> there's no distinction between readdir and read in plan 9.
> >=20
> > Forgot about that. Still, you can't chdir() into an inode that doesn't
> > indicate being a directory. And the bytestream returned by
> > read(SOME_DIRECTORY) is fixed-format and doesn't provide any space for
> > free- form bytestream.
>=20
> i don't think that helps you.
>=20
> under plan 9, reading a directory is this:
>=20
> translate(read(open(dir)))
>=20
> i don't see how you can make read(open(dir)) return
> something different.

no contest there. I'd like to have a look at the handling of multiple forks=
 of=20
files in other OSes (from userland perspective).

=20
> oh yes, maintaining the usual semantics for cp becomes tricky.
>=20
> mkdir z
> cp x.c z
>=20
> do i mean to write x.c to z itself, or to a new file within z?

I have only linux experience with cp -- and this usage pattern always irked=
=20
me. Theoretically, it's prone to race condition: suppose somebody removes t=
he=20
dir `z' right between `mkdir' and `cp'. You `cp x.c z', and end up with `z'=
=20
file. It's unpredictable, unless you're on a single-process OS (yuck).

I'd rather see `cp x.c z' always duplicates bytestream from x.c into z's=20
bytestream, whatever `z' happens to be right now. Creating `z' in the proce=
ss,=20
if not present. If you want to have `z/x.c', you'd state that explicitly as=
=20
`cp x.c z/x.c'.

nb., with the current semantics you *could* say `cp x.c z/' to be unambiguo=
us=20
you want to create a child of `z', but it seems to be common not to use=20
trailing slash unless 100% necessary.


=2D-=20
dexen deVries

[[[=E2=86=93][=E2=86=92]]]

> how does a C compiler get to be that big? what is all that code doing?

iterators, string objects, and a full set of C macros that ensure
boundary conditions and improve interfaces.

ron minnich, in response to Charles Forsyth

http://9fans.net/archive/2011/02/90