Simple HowTo

Gene Poole gene.poole at
Wed Dec 19 03:14:35 UTC 2007

Craig wrote:

> Gene Poole wrote:
> > Hello everyone!
> >
> > I started this thread so let me see if I can clarify my point of view.
> >
> > Initially I asked if there was a simple HowTo for Apache HTTPD, Apache
> > Tomcat, Sun Java, and mod_jk.  I did this because I found on the
> > at least 6 different ways to make this work, but all were incomplete in
> > way or another. Also, I could not find a way to combine them. The
> > was for me to use the RPM packages provided by Fedora with the
exception of
> > the Sun Java where I was on my own. And why not use IcedTea?  Why not
> > the full complete Java? I had not gotten to my need for Eclipse and the
> > fact I run Oracle 10g on this machine.
> ----
> the concept of IcedTea is to have the free-java implementation aka gcj
> peacefully co-exist with another java install (i.e. sun's java). This is
> because the gcj implementation is used as prerequisite for other
> ----

But the Sun Java is free.  It is just under a license that doesn't allow it
to be bundled with a Fedora distro. IcedTea may be a great thing, but
Oracle ships with a JVM for itself.  For Oracle 10g it's 1.4.2. Jedit is a
Java program that doesn't seem to want to work unless it can find
($JAVA_HOME) certain jar files in the Sun Java Classpath.

> >
> > Here are my concerns about that:
> >       I use LVM, but I don't believe that is normal.  I think that when
> >       install is done, the default configuration is used (i.e. /boot;
> >       /home; and swap) regardless of the size of the hard drive(s).
> >       The latest releases (Fedora 7 and 8) have moved away from the
> >       IDE constructs (up to 15 partitions) to a more SCSI construct (up
> >       6(?) partitions). So all older definitions where we learned that
> >       of the better installations defined specific file systems and
> >       points now must move to a more M$ C: drive mentality.  If I want
> >       avoid this I must use LVM so that I can define file systems for
> >       entire tree.
> ----
> - you don't have to use LVM...the choice is made at install.

I think you will find that if you select the 'Standard' install where the
installer creates the file systems, it doesn't use LVM but standard
partitioning. I am speaking of a clean machine with no pre-existing
operating system.

> - hard drive constructs, as you call them are no different in F7/F8 than
> earlier versions of Fedora/RHEL with 'SCSI' hard can still
> have 16 partitions, the issue is that  if you use fdisk to create them,
> there can only be 4 primary partitions and typically the 4th primary
> partition is made as an extended partition which is then partitioned
> within. I'm surprised given all of your history and research you hadn't
> grasped that yet.

But they are different from what they were with Fedora Core 6 - a big
difference.  It no longer uses that standard IDE constructs, but makes
everything look like SCSI (i.e. that are no longer hda - hdz drives only
SCSI drive types).  Under IDE I could have many more partitions per hard
drives than you can with SCSI. I thought you would have grasped that by
now. I know this because I attempted to do a install of Fedora 7 on a
machine that was running Fedora Core 6 and I wanted to do it without
re-partitioning the hard drives. No good, I had to start using LVM because
I needed the equivalent if 17 partitions.

> - none of this has any connection whatsoever to an 'M$ C: drive
> mentality'. In fact, the one with the 'M$ C: drive mentality' is you
> because you can't seemingly grasp all of the versatility allowed to you
> with the 'virtual' features of LVM. You can have as many LVM 'volumes'
> as you wish and declare the mount points where you wish. You can have
> /home as a separate LVM volume if you wish...that's about as anti 'M$ C:
> drive mentality' as it could possibly get.
> ----

I'm not sure I can get you to see my point.  The absolute easiest way to
install Fedora anything is to let the installer define the file systems.
And the easiest way for the installer to do that is to build a tree that is
based on the four partitions I mentioned above.

> >       I have no doubt that the RPMs contain all that are needed, except
> >       can no longer control where it goes.  You don't tell me where
> >       going to go ahead of time so I can make a file system large
> >       and named correctly (more C: drive mentality).  And I need to
> >       this ahead of time so I can do the correct thing while in Disk
> >       In the FHS, where does it say that Apache HTTPD, Apache Tomcat,
> >       mod_jk, and the Sun Java should go?
> ----
> You could download the source RPMs and edit the 'spec' files to control
> where the stuff gets installed but of course, that would take
> knowledge...research
> ----

This is another point I tried to make.  Take a Windows user attempting to
use/migrate to Linux and try and see their frustration.  They don't want to
know more than the minimum necessary to get it to work.  I'm beyond that (I
hope).  But now you're telling me that I should build a rpm rather than use
the prepackaged tar.gz.  That is a pretty wild statement.  And remember
this was not the case until Fedora 7.  No problem using tar.gz packages in
Fedora 6. What did they change and why?

> > Other Points:
> >       /opt and /usr/local are going to be defined regardless if you
> >       built a separate file system for them or not.  They are a part of
> >       standard file hierarchy.
> ----
> if they didn't exist, how would you mount a filesystem there?
> can you demonstrate cluelessness any more elegantly?
> ----

I think you've forgotten that a mount point is nothing more than a
sub-directory.  If I only defined a partition or logical volume for /usr,
and then if I did a ls -ltr you would see /usr/local.  If I only defined a
partition or logical volume for / (root), and then I did a ls -ltr you
would see /opt somewhere in the listing. The complete tree is defined
regardless if it is on a separate partition of just a sub-directory below
root defined by the OS.  So I'm not so clueless as you might think.

> >       Yum doesn't know nor should it know about software installed
> >       of it's environment.  One great piece of software is jedit and
> >       is no RPM for it - are you saying I shouldn't use it? There are a
> >       of good packages out there that aren't packaged in RPM format
> >       is a java program). The defacto standard for packages is .tar.gz
> >       I've used yum when it was only on the Yellowdog distro (yum =
> >       yellowdog update manager).  I started in Linux when there was no
> ----
> but Sun does distribute jre/jdk in rpm format. I don't have a clue how
> they distribute a jar file perhaps?
> ----

jedit is a jar file and I usually install it in my /home directory. Yes,
the java rpm does exist.  But, google to find the instructions for
installing the SUN JDK 1.6 on Fedora 7 or Fedora 8 and the majority state
not to use the RPM.  Why, because those instructions want you to place the
JDK in /opt.  I like my Java in /usr/java; I like my tomcat in /usr/tomcat;
I like my apache in /usr/apache; I like my eclipse in /usr/eclipse; I like
my JBoss in /usr/JBoss; I like my oracle in /usr/oracle. And if there is a
tar.gz package that works and is available, why should I have to build a
rpm?  Explain that and if the explanation is good enough, I'll start
building RPMs. If you install Oracle 10g or 11g database on Fedora 8, you
now have to do a 'sed' on the JRE to get it to work.  Why, because
something in the system has changed to the point where the standard java
JRE supplied no longer works.  That tells me there is something in Fedora 8
that will never make it to RHEL.  And remember that's what this is - a
staging ground for RHEL.

> >
> > Resources:
> >
> ----
> why not read the whole thing instead of just /usr/share
> There's a lot of information there.

I have read it. And besides, I'm not the person that mentioned /usr/share,
it was someone else.

> Craig

Gene Poole
gene.poole at

More information about the users mailing list