Well, I figured out that one of my issues was that the option to
timezone had been replaced with (which I've not yet found documented)
and that it wasn't backward compatible (even though things like the "yes"
argument to "zerombr" are non-fatal warnings). Hmmm.
Anyway, was wondering about a few other things
For instance, on "network", can you specify this command multiple times,
as in:
network
network
for a laptop with both wired and wireless interfaces?
And what about adding:
network
Why? Well, when I worked at Cisco we had some users that worked over
VPN where the overhead for IPSEC/GRE tunneling reduce the maximum
UDP data size (for TFTP, etc) to 1400 bytes. A proper client should specify
that it wanted the MaxMTU option from DHCP and then set the interface
appropriately, but not all do (PIXE comes to mind). And then there are
static IP addresses, of course
Being able to netboot a PC and image it without having to drive 200 miles
to come into the nearest corporate office would have been cool.
Another thing I was thinking about was being able to perform simple edits
of the contents of /etc/sysconfig files. I've written the following
script to
give an example of this. It has drawbacks, like not supporting
subdirectories,
but that's easily fixed
The way I use it right now is:
%post
/inst-helper/sysconfig.pl -b.orig 'init/GRAPHICAL=no'
'network/+NETWRKING_IPv6=no'
cat >/etc/modprobe.conf <<__EF__
# turn off IPv6
alias net-pf-10 off
alias ipv6 off
__EF__
Does this seem like it (the functionality of the sysconfig.pl script)
would be a
useful extension to the parameter section of the Kickstart config file
(i.e. make
it a built-in command "sysconfig" just like "xconfig" or "timezone")?
Thanks,
-Philip
Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Tue, 2006-05-09 at 11:34 -0600, Philip Prindeville wrote:
Well, I figured out that one of my issues was that the option to
timezone had been replaced with (which I've not yet found documented)
and that it wasn't backward compatible (even though things like the "yes"
argument to "zerombr" are non-fatal warnings). Hmmm.

In specific cases like this, if you file bugs against pykickstart, Chris
is quick about fixing them ;)

For instance, on "network", can you specify this command multiple times,
as in:

network
network

for a laptop with both wired and wireless interfaces?

That used to work. Note that it starts to be a little bit less relevant
as we move towards a world using NetworkManager.

And what about adding:

network

Could be easily added, just hasn't been requested before.

Another thing I was thinking about was being able to perform simple edits
of the contents of /etc/sysconfig files. I've written the following
script to give an example of this. It has drawbacks, like not supporting
subdirectories, but that's easily fixed

This is pretty trivially done with sed in %post I don't see a need to
further complicate the parameters which are already there. Every new
command adds an increased testing burden and the amount of things that
people want to do is infinite. Which is why there is the infinite
flexibility of %post available.

Jeremy

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Tue, 2006-05-09 at 13:57 -0600, Philip Prindeville wrote:
Jeremy Katz wrote:
Tue, 2006-05-09 at 11:34 -0600, Philip Prindeville wrote:
>Well, I figured out that one of my issues was that the option to
>timezone had been replaced with (which I've not yet found documented)
>and that it wasn't backward compatible (even though things like the "yes"
>argument to "zerombr" are non-fatal warnings). Hmmm.
>
>
In specific cases like this, if you file bugs against pykickstart, Chris
is quick about fixing them ;)

>For instance, on "network", can you specify this command multiple times,
>as in:
>>
>network
>network
>>
>for a laptop with both wired and wireless interfaces?
>
That used to work. Note that it starts to be a little bit less relevant
as we move towards a world using NetworkManager.

I'm thinking that in the case where you're installing to a barebones
headless system (like an EPIA with an Ethernet and a wireless interface,
or other embedded device or server) that NetworkManager might be
irrelevant.

NetworkManager should (long-term) be equally relevant for handling *all*
network connectivity. Having multiple methods of network configuration
is the path to madness.

Why did it stop working?

AFAIK, it still does. If it doesn't, it's a bug as a result of the
refactoring of the kickstart code in FC5.

>Another thing I was thinking about was being able to perform simple edits
>of the contents of /etc/sysconfig files. I've written the following
>script to give an example of this. It has drawbacks, like not supporting
>subdirectories, but that's easily fixed
>
This is pretty trivially done with sed in %post I don't see a need to
further complicate the parameters which are already there. Every new
command adds an increased testing burden and the amount of things that
people want to do is infinite. Which is why there is the infinite
flexibility of %post available.

Hmmm , how about coming up with Anaconda helper scripts then
and adding them to the release?

I have a script, for instance, that tickles the right options for creating
printers via CUPS using the lpadmin interface, which was non-trivial
to discover and isn't very well documented. It might be something that
could be useful to have in a generic toolkit.

I think the way to go here is actually continuing to put more content on
the wiki[1] on the kickstart snippets page[2] and ensuring the docs in
the package and for kickstart have a pointer to there. As I really
don't want to be in the business of maintaining a lot of little helper
scripts

Jeremy

[1]
[2]

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Jeremy Katz wrote:
Tue, 2006-05-09 at 13:57 -0600, Philip Prindeville wrote:

>Jeremy Katz wrote:
>
Tue, 2006-05-09 at 11:34 -0600, Philip Prindeville wrote:

Well, I figured out that one of my issues was that the option to
timezone had been replaced with (which I've not yet found documented)
and that it wasn't backward compatible (even though things like the "yes"
argument to "zerombr" are non-fatal warnings). Hmmm.

In specific cases like this, if you file bugs against pykickstart, Chris
is quick about fixing them ;)

For instance, on "network", can you specify this command multiple times,
as in:

network
network

for a laptop with both wired and wireless interfaces?

That used to work. Note that it starts to be a little bit less relevant
as we move towards a world using NetworkManager.

>I'm thinking that in the case where you're installing to a barebones
>headless system (like an EPIA with an Ethernet and a wireless interface,
>or other embedded device or server) that NetworkManager might be
>irrelevant.
>
>
NetworkManager should (long-term) be equally relevant for handling *all*
network connectivity. Having multiple methods of network configuration
is the path to madness.

My point, which I didn't make very well, is that you might find yourself
in an environment where there is no GUI support.


>Why did it stop working?
>
>
AFAIK, it still does. If it doesn't, it's a bug as a result of the
refactoring of the kickstart code in FC5.

I'll try it again. It failed last time, and I didn't investigate.
I suspect it
was because the madwifi RPM hadn't been installed, nor had the kernel
module been modprobe'd so it wasn't seeing the ath0 interface (would
be nice if there was a way to force Anaconda to go ahead and configure
an interface, even when ifconfig -a doesn't report it as being present
would be useful for removable interfaces and PCMCIA cards, etc).

Another thing I was thinking about was being able to perform simple edits
of the contents of /etc/sysconfig files. I've written the following
script to give an example of this. It has drawbacks, like not supporting
subdirectories, but that's easily fixed

This is pretty trivially done with sed in %post I don't see a need to
further complicate the parameters which are already there. Every new
command adds an increased testing burden and the amount of things that
people want to do is infinite. Which is why there is the infinite
flexibility of %post available.

>Hmmm , how about coming up with Anaconda helper scripts then
>and adding them to the release?
>>
>I have a script, for instance, that tickles the right options for creating
>printers via CUPS using the lpadmin interface, which was non-trivial
>to discover and isn't very well documented. It might be something that
>could be useful to have in a generic toolkit.
>
>
I think the way to go here is actually continuing to put more content on
the wiki[1] on the kickstart snippets page[2] and ensuring the docs in
the package and for kickstart have a pointer to there. As I really
don't want to be in the business of maintaining a lot of little helper
scripts

Jeremy

[1]
[2]
--

Fair enough.
-Philip

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Tue, 2006-05-09 at 14:39 -0600, Philip Prindeville wrote:
Jeremy Katz wrote:
Tue, 2006-05-09 at 13:57 -0600, Philip Prindeville wrote:
>Jeremy Katz wrote:
Tue, 2006-05-09 at 11:34 -0600, Philip Prindeville wrote:
NetworkManager should (long-term) be equally relevant for handling *all*
network connectivity. Having multiple methods of network configuration
is the path to madness.

My point, which I didn't make very well, is that you might find yourself
in an environment where there is no GUI support.

The fact that NetworkManager doesn't currently function on a system-wide
basis without a GUI is a bug that needs fixing (it's one of the
prerequisites for enabling NetworkManager by default and why that didn't
happen for FC5). But that's veering towards off-topic for here :)

Jeremy

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Jeremy Katz wrote:
Tue, 2006-05-09 at 11:34 -0600, Philip Prindeville wrote:

>Well, I figured out that one of my issues was that the option to
>timezone had been replaced with (which I've not yet found documented)
>and that it wasn't backward compatible (even though things like the "yes"
>argument to "zerombr" are non-fatal warnings). Hmmm.
>
>
In specific cases like this, if you file bugs against pykickstart, Chris
is quick about fixing them ;)


>For instance, on "network", can you specify this command multiple times,
>as in:
>>
>network
>network
>>
>for a laptop with both wired and wireless interfaces?
>
>
That used to work. Note that it starts to be a little bit less relevant
as we move towards a world using NetworkManager.

I'm thinking that in the case where you're installing to a barebones
headless system (like an EPIA with an Ethernet and a wireless interface,
or other embedded device or server) that NetworkManager might be
irrelevant.

Why did it stop working?


>And what about adding:
>>
>network
>
>
Could be easily added, just hasn't been requested before.

I'll file a defect

>Another thing I was thinking about was being able to perform simple edits
>of the contents of /etc/sysconfig files. I've written the following
>script to give an example of this. It has drawbacks, like not supporting
>subdirectories, but that's easily fixed
>
>
This is pretty trivially done with sed in %post I don't see a need to
further complicate the parameters which are already there. Every new
command adds an increased testing burden and the amount of things that
people want to do is infinite. Which is why there is the infinite
flexibility of %post available.

Hmmm , how about coming up with Anaconda helper scripts then
and adding them to the release?

I have a script, for instance, that tickles the right options for creating
printers via CUPS using the lpadmin interface, which was non-trivial
to discover and isn't very well documented. It might be something that
could be useful to have in a generic toolkit.
-Philip

Jeremy

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Jeremy Katz wrote:
Tue, 2006-05-09 at 11:34 -0600, Philip Prindeville wrote:

>Well, I figured out that one of my issues was that the option to
>timezone had been replaced with (which I've not yet found documented)
>and that it wasn't backward compatible (even though things like the "yes"
>argument to "zerombr" are non-fatal warnings). Hmmm.
>
>
In specific cases like this, if you file bugs against pykickstart, Chris
is quick about fixing them ;)

I've filed bugs for the retro-compatibility issue (191229) and
for the
desired feature (191328).
-Philip

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Would it make sense to add a "" (or , or whatever)
mode to anaconda to syntax-check a ks.cfg file and see if it's valid?
Sure would save a lot of time
-Philip
Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com

Would it make sense to add a "" (or , or whatever)
mode to anaconda to syntax-check a ks.cfg file and see if it's valid?

Sure would save a lot of time

/usr/bin/ksvalidator in the pykickstart package will do this for you.
- Chris

Anaconda-devel-list mailing list
Anaconda-devel-list (AT) redhat (DOT) com