[gentoo-releng] Re: Feature Requests for 2006.1

[gentoo-releng] Re: Feature Requests for 2006.1

[Mikey]

[-- Attachment #1: Type: text/plain, Size: 1509 bytes --]

Chris Gianelloni wrote:

> into the 2006.1 profiles.  I am looking for suggestions for USE flags to
> add to these two profiles, so feel free to make suggestions.  Please
> cite some reasoning for why you think each USE flag you recommend should
> be either enabled or disabled.  The profiles are at
> profiles/default-linux/x86/dev/2006.1/desktop and
> profiles/default-linux/x86/dev/2006.1/server for you to peruse.  They
> are completely functional profiles at this time.

In the servers profile...

logrotate would be nice for obvious reasons on servers.  

chroot might be nice, as long as it is not too invasive (requires lots of
extra configuration of the packages that utilize it).

My main concern is not really what USE flags need to be added as opposed to
what USE flags might need to be removed.  In my opinion a generic server
profile needs to be as generic as possible.  For example, cups foomatic gpm
and ldap from dev/2006.1/make.defaults should not go into a generic server
profile because in some cases they make significant differences in how
subsequent packages will be configured - samba and apache2 for examples.

None of my servers have pointing devices, gpm is not only useless in this
situation, it introduces additional unnecessary maintenance.  mailwrapper
is another example of something that only serves to give me headaches ;)

I noticed you have STAGE1_USE="nptl nptlonly", does that mean that the CHOST
will need to be changed in stage1 tarballs?

[-- Attachment #2: Type: application/pgp-signature, Size: 191 bytes --]

Re: [gentoo-releng] Re: Feature Requests for 2006.1

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 2676 bytes --]

On Wed, 2006-03-15 at 09:34 -0600, Mikey wrote:
> logrotate would be nice for obvious reasons on servers

The only package that uses this correctly is squid, so I'd prefer not to
add it.

> chroot might be nice, as long as it is not too invasive (requires lots of
> extra configuration of the packages that utilize it).

This would be a bit much.

> My main concern is not really what USE flags need to be added as opposed to
> what USE flags might need to be removed.  In my opinion a generic server
> profile needs to be as generic as possible.  For example, cups foomatic gpm
> and ldap from dev/2006.1/make.defaults should not go into a generic server
> profile because in some cases they make significant differences in how
> subsequent packages will be configured - samba and apache2 for examples.

I would remove gpm, but the others are very unlikely.  The purpose here
would be to create something that is actually usable as a default.
You're more than welcome to customize it yourself, and are expected to
do so.  Just like how the default USE under the 2006.0 and
2006.1/desktop profiles have lots of things some people won't want (both
gnome and kde, for example), it is intended to enable support that most
people would want, while still remaining somewhat minimal.

> None of my servers have pointing devices, gpm is not only useless in this
> situation, it introduces additional unnecessary maintenance.  mailwrapper
> is another example of something that only serves to give me headaches ;)

Again, just because none of *your* servers do not have pointing devices
does not make it an accurate general statement.  My main goal here is to
keep all of the desktop USE flags out of the profile.  In this case, I
can definitely see a use for gpm on a server, unlike gnome or xmms.

> I noticed you have STAGE1_USE="nptl nptlonly", does that mean that the CHOST
> will need to be changed in stage1 tarballs?

Actually, I'm building this currently thinking that glibc 2.4 would be
used, which is only nptl.  I am not going to be building another set of
no-nptl stages on x86.  The 2006.0 stages will be considered *it* for
building on any non-nptl system without using hardened stages.

Of course, any and all of this is likely to change after further
discussion with solar and the rest of the hardened/server/infra guys.

Honestly, I don't want people to focus on the server profile as much as
what really concerns *me* which is the desktop setup that will be used
for building the next LiveCD set.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[gentoo-releng] Re: Re: Feature Requests for 2006.1

[MIkey]

Chris Gianelloni wrote:

>> None of my servers have pointing devices, gpm is not only useless in this
>> situation, it introduces additional unnecessary maintenance.  mailwrapper
>> is another example of something that only serves to give me headaches ;)
> 
> Again, just because none of *your* servers do not have pointing devices
> does not make it an accurate general statement.  My main goal here is to
> keep all of the desktop USE flags out of the profile.  In this case, I
> can definitely see a use for gpm on a server, unlike gnome or xmms.

I would be willing to bet that most people who run servers don't plug mice
into them and don't need them.

Last time I removed gpm from a stage3 it became unusable because the gpm
libraries were linked into coreutils (ls), even though gpm is not in IUSE
for coreutils.  This is a perfect example of why I prefer a minimal base
setup - to limit headaches such as this and the need to update packages
with possible vulnerabilities in the future even though I don't use them.
 
> Honestly, I don't want people to focus on the server profile as much as
> what really concerns *me* which is the desktop setup that will be used
> for building the next LiveCD set.
 
So, is your request for comments on server profiles now cancelled?

-- 
gentoo-releng@gentoo.org mailing list