Mailinglist Archive: opensuse-factory (422 mails)
| < Previous | Next > |
Re: [opensuse-factory] Plan for 11.2?
- From: "Matt Sealey" <matt@xxxxxxxxxxxxxx>
- Date: Thu, 15 Jan 2009 12:34:07 -0600
- Message-id: <b5e2fc790901151034k1ae33533xae01a08468649073@xxxxxxxxxxxxxx>
On Thu, Jan 15, 2009 at 11:21 AM, Rob OpenSuSE
<rob.opensuse.linux@xxxxxxxxxxxxxx> wrote:
We have a very very good SUSE guy at Genesi (Peter Czanik) who did
these tests. He runs through Factory, he tests every beta, and
compares them to Fedora and Ubuntu too. SUSE uses less memory than the
other two, but it is still at 11.1 using far more memory than it used
to.
.. right.
No shit, sherlock.
Because I left the installer on Automatic Configuration and somehow it
decided to install it, and enable it. I do not have a printer
connected on install, let alone an HP one, let alone a multifunction
one. This is a decision made by the installer, automatically.
Why should I be asking myself why it is installed on my system? I
should be asking the SUSE developers. But, if you were paying any
attention, you'd notice *this is exactly what I asked for the past 3
days on this list*. Once I have enough information on why this may be
happening - because I cannot see or find the procedure in the
installer or any other script or package or pattern which makes this
happen - I can file a bug report on it because it is obviously
absolutely the wrong thing to do.
With regards to cups, this is something I could live with; as a result
of Avahi being enabled on boot, which autodetects printers by mDNS, it
needs something to attach them to, but first of all, I don't use
Avahi, so I usually disable it. There is, however, no way to really do
this cleanly except pop into a services management tool and manually
kill off two services or uninstall it entirely. I would expect, then,
that cups is not started since it has nothing dependent on it, no
printers set up, etc.. I would also expect that improvements can be
made such that Avahi - while detecting printers which may or more
prudently may NOT exist on a network - does not actually require the
cups daemon to be running idle until a multicast notification for a
printer comes in, at which point it will notify something that can and
will start up a cups daemon. In such a case, it would be started
exactly when it's needed and not before.
I know there are problems with FUSE - I do not use it myself so I
disable it. If you start fusermount when the kernel module is not
loaded, it WILL load it by itself, which is admirable behaviour.
However there are quirks with permissions which mean that you cannot
load a kernel module as a user (thus defeating the object of FUSE
entirely IMO) plus some permissions on /dev/fuse are not set
correctly. Thus, it was decided it needs to be started on boot to let
people actually use it. I saw distributions looking at this bug in
January 2006 and now, some 3 years later, the "let's not break fuse
too much" fix is still in effect. Loading the module at boot is the
simplest hackiest way to get it working. There must be a better way
but since it "works" for everyone, nobody is paying attention,
regardless of whether it is the best solution or not.
It is not a performance issue (although the huge script for loading
rpcbind, and loading rpcbind itself takes time on boot) but a sort of
a moral, ethical issue. Should we be starting services which are
strictly dependencies on other services, for the sake of having them
around just in case they're needed?
FUSE gets used by GNOME and KDE VFS, cups may need to be around for
Avahi, but right now nothing is installed in the default configuration
(through automatic configuration, the default way to install) that
requires rpcbind and if you didn't set up an LVM disk and none were
detected in the installer, device-mapper isn't needed either (I don't
see why you can't start the lvm and device-mapper stuff when the tools
are used as fusermount does.. but oh well). They needn't be started at
all UNTIL the point when you actually need it. In the case of
something like Avahi, isn't this *WAY* past the stage that gdm has
loaded on most default configurations? It needn't be blocking gdm.
--
Matt Sealey <matt@xxxxxxxxxxxxxx>
Genesi, Manager, Developer Relations
--
To unsubscribe, e-mail: opensuse-factory+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse-factory+help@xxxxxxxxxxxx
<rob.opensuse.linux@xxxxxxxxxxxxxx> wrote:
2009/1/15 Matt Sealey <matt@xxxxxxxxxxxxxx>:
On Thu, Jan 15, 2009 at 10:42 AM, Rob OpenSuSE
<rob.opensuse.linux@xxxxxxxxxxxxxx> wrote:
This says more about you. I looked and made comparisons, I report
that I do not see general bloat increase from 10.3 to 11.1 and give
some expamples of lowered memory consumption.
We have a very very good SUSE guy at Genesi (Peter Czanik) who did
these tests. He runs through Factory, he tests every beta, and
compares them to Fedora and Ubuntu too. SUSE uses less memory than the
other two, but it is still at 11.1 using far more memory than it used
to.
My message is you need to look elsewhere, not these general statements
for solutions.
.. right.
I do believe you have problems, but they don't affect me, nor well
specified systems that have more RAM.
No shit, sherlock.
modules (boot.lvm or boot.dmraid) and/or tools. Cups and HPLIP only
*need* to be started as explicit dependencies on *actually having a
printer* (it does not affect udev hotplug of USB printers that cups is
Desktop OS, server most ppl need to print. HPLIP as I told you before
are not installed on my system, why don't you ask yourself why it is
on yours?
Because I left the installer on Automatic Configuration and somehow it
decided to install it, and enable it. I do not have a printer
connected on install, let alone an HP one, let alone a multifunction
one. This is a decision made by the installer, automatically.
Why should I be asking myself why it is installed on my system? I
should be asking the SUSE developers. But, if you were paying any
attention, you'd notice *this is exactly what I asked for the past 3
days on this list*. Once I have enough information on why this may be
happening - because I cannot see or find the procedure in the
installer or any other script or package or pattern which makes this
happen - I can file a bug report on it because it is obviously
absolutely the wrong thing to do.
With regards to cups, this is something I could live with; as a result
of Avahi being enabled on boot, which autodetects printers by mDNS, it
needs something to attach them to, but first of all, I don't use
Avahi, so I usually disable it. There is, however, no way to really do
this cleanly except pop into a services management tool and manually
kill off two services or uninstall it entirely. I would expect, then,
that cups is not started since it has nothing dependent on it, no
printers set up, etc.. I would also expect that improvements can be
made such that Avahi - while detecting printers which may or more
prudently may NOT exist on a network - does not actually require the
cups daemon to be running idle until a multicast notification for a
printer comes in, at which point it will notify something that can and
will start up a cups daemon. In such a case, it would be started
exactly when it's needed and not before.
I know there are problems with FUSE - I do not use it myself so I
disable it. If you start fusermount when the kernel module is not
loaded, it WILL load it by itself, which is admirable behaviour.
However there are quirks with permissions which mean that you cannot
load a kernel module as a user (thus defeating the object of FUSE
entirely IMO) plus some permissions on /dev/fuse are not set
correctly. Thus, it was decided it needs to be started on boot to let
people actually use it. I saw distributions looking at this bug in
January 2006 and now, some 3 years later, the "let's not break fuse
too much" fix is still in effect. Loading the module at boot is the
simplest hackiest way to get it working. There must be a better way
but since it "works" for everyone, nobody is paying attention,
regardless of whether it is the best solution or not.
not running!). rpcbind only *NEEDS* to be started if a service using
rpcbind/portmap is started too.
Agreed, but it's not causing performance problems.
It is not a performance issue (although the huge script for loading
rpcbind, and loading rpcbind itself takes time on boot) but a sort of
a moral, ethical issue. Should we be starting services which are
strictly dependencies on other services, for the sake of having them
around just in case they're needed?
FUSE gets used by GNOME and KDE VFS, cups may need to be around for
Avahi, but right now nothing is installed in the default configuration
(through automatic configuration, the default way to install) that
requires rpcbind and if you didn't set up an LVM disk and none were
detected in the installer, device-mapper isn't needed either (I don't
see why you can't start the lvm and device-mapper stuff when the tools
are used as fusermount does.. but oh well). They needn't be started at
all UNTIL the point when you actually need it. In the case of
something like Avahi, isn't this *WAY* past the stage that gdm has
loaded on most default configurations? It needn't be blocking gdm.
--
Matt Sealey <matt@xxxxxxxxxxxxxx>
Genesi, Manager, Developer Relations
--
To unsubscribe, e-mail: opensuse-factory+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse-factory+help@xxxxxxxxxxxx
| < Previous | Next > |