Re: [9fans] changes in 9load

[Fco.J.Ballesteros <nemo@plan9.escet.urjc.es>]

[-- Attachment #1: Type: text/plain, Size: 1050 bytes --]

I wanted the variables to start fossil/venti without having to
recompile the kernel. I also would prefer to put the conf in the
disks they're using (In fact I think I suggested it some time ago).
The thing is that I wouldn't like to change fossil/venti, because I
don't know them that well, and I'd prefer that change to be well-done
(for example, I'd expect the changed fossil/venti to be able to use
partitions from old ones).

But in the mean time, I think I'll keep the support for multiline
vars (which I'll be using just in our file server). I'm quite tired
of recompiling the kernel because of a change in the configuration.

I think the same can be the case for others, althought of course
that means that their fs plan9.ini would be `mandatory' as well.

But it's likely we all agree that once there's no need to use those
configuration files, multiline vars would become useless.

Regarding the other change, it seems better to parse the arguments once
and pass them already parsed to the kernel.

thanks for your reply

[-- Attachment #2: Type: message/rfc822, Size: 3390 bytes --]

From: "Russ Cox" <rsc@plan9.bell-labs.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] changes in 9load
Date: Wed, 21 May 2003 11:22:24 -0400
Message-ID: <378f47e82d876408e51c03e6eb666d80@plan9.bell-labs.com>

[The "I might.  I will check tonight." was a response to a
different, private email from Nemo.  Sorry.  Here's
a real response to make up for it.]

I'd rather not see the multiline environment variable hack
go into plan9.ini.  (Though if it did, I'd name the sections [$var]
instead of [!var].)

Plan9.ini used to be this really complex mess that you had to
fiddle with quite a bit to get your system to boot.  Now it's this
really complex mess that you only have to fiddle with a small
amount to get your system to boot.  A plan9.ini for a typical
configuration only needs to set mouseport, vgasize, and monitor.
All three of these could be stored in the file system proper instead,
making plan9.ini completely optional.  I haven't done this because
I'd rather see the variables go away entirely.  If we autodetect the
mouse and probe the video card for monitor and a good vgasize,
that gets us most of the way there.  If aux/vga can handle resizing
the screen on the fly, we're all the way there.

There will always be a plan9.ini.  There has to be some way
to tell the kernel things that it doesn't know how to learn for itself.
But I want to see plan9.ini get more and more optional.
Unless you're doing something complicated, you shouldn't
have to know about it.

One eventual goal for fossil is to replace kfs.  This means
everyone will be running fossil (with or without venti).  Needing
to put configuration info in plan9.ini suddenly makes plan9.ini
a lot less optional.  Thus I am fairly opposed to having it there.
Instead, I would like to see the configuration information for
venti and fossil put at the beginning of their disks, so that
you can just tell it the disk and you're off and running.  With
some sensible conventions naming the disks we could boot
an IDE system with no plan9.ini at all.

Long ago, plan9.ini had syntax for defining multiline
environment variables.  It was dropped at one point because
it wasn't found to be generally useful.  I hope this continues
to be the case.  If we need to store that much in an environment
variable, it means we've built the system in such a way that
it requires too much configuration.

Russ