Linux-Misc Digest #474, Volume #20 Thu, 3 Jun 99 08:13:08 EDT
Contents:
Re: log a user's inbound/outbound mail? (Sir Hoagy of the Marshlands)
Re: Operating systems (Richard Kulisz)
Re: RH 6.0 and .bash_profile (Jon Edwards)
Re: Decent Partition Sizes?? (Azfar Kazmi)
Re: Does Java run well on Linux? (Stephan Houben)
Re: Decent Partition Sizes?? (Azfar Kazmi)
Kernel Hacking ([EMAIL PROTECTED])
scsi error on hd ("Jan Rockstedt")
Re: INN news server (Carl Fink)
Re: Is This Illegal? (Carl Fink)
----------------------------------------------------------------------------
From: Sir Hoagy of the Marshlands <[EMAIL PROTECTED]>
Subject: Re: log a user's inbound/outbound mail?
Date: Wed, 02 Jun 1999 19:33:55 -0400
You *could* set up cron to scan his mailbox every 30
seconds or so, and if its contents are greater than
0k, copy the contents (not move, but copy) to a secure
directory.
This -shouldn't- eat up a lot of system resources, even
on a medium-to-large multi-user system; however, it doesn't
sound, even to me, to be the best method of accomplishing
your goals.
This method *assumes* he isn't using pop or imap to fetch his mail;
rather, this assumes his email stays in /var/mail/<username>
If so, this method should be effective, even if he gets his
mail very often.
It's a suggestion - I'm more than eager to hear from others
who have a better idea (so I can steal it, of course!)
>>Matt<<
------------------------------
From: [EMAIL PROTECTED] (Richard Kulisz)
Crossposted-To: comp.os.ms-windows.advocacy,comp.os.linux.advocacy,gnu.misc.discuss
Subject: Re: Operating systems
Date: 3 Jun 1999 07:33:28 GMT
In article <[EMAIL PROTECTED]>,
The Ghost In The Machine <[EMAIL PROTECTED]> wrote:
>On 2 Jun 1999 04:03:31 GMT, Richard Kulisz <[EMAIL PROTECTED]> wrote:
>_Design Patterns: Elements of Resuable Object-Oriented Software_,
>by Erich Gamma, Richard Helm, Ralph Johnson, and John Vlissides.
>Published by Addison-Wesley.
That's the one. There's so much hype around Patterns it's made me sick.
>It's a work for the clarification, proper use, and implementation
>of various patterns in OO software. I'm not sure it's wonderfully
>written, but it does contain some useful info. :-)
>
>(I know Abstract Factories, Chains of Responsibility,
>Visitors, and Singleton classes might be useful right off the
>top of my head...)
>No, but cycles in themselves are not harmful. It's merely a redundancy
>in the information of the system; they generate an infinite number
>of pathname representations for a given object, but are otherwise
>harmless unless disconnected from the root -- in which case, they're
>just garbage to be collected.
I never thought of that; great, another reason why cycles are bad.
I don't want to have to check accessbility before garbage collecting
hard disk space (which is a huge overhead as it is) and I especially
don't want to make root a special case.
>The algorithm given below does that very easily. It's not a problem.
You're right, avoiding infinite loops is not a problem.
>*What* parent links? That's redundant information; the proper way
>of doing this is to set up a separate datastructure, and be able
>to search for it two ways. The data structure is (P,C,iname);
>P and C are internal Inode numbers, and iname is the name of
>C inside of P.
I bet you like Prolog a lot, am I right?
I don't think setting up a separate data structure is an efficient
way to do it, especially since I don't plan to export inode numbers
the way Unix does; they're internal and so accessible only to internal
methods. I also want to be able to add inodes whenever necessary.
>It's a very common method of handling relational database
>many-many relationships (in this case, an Inode can have many
>Inode children, and Inode children can have many Inode parents;
>one might consider it a self-cross).
Unfortunately, I haven't read anything on databases.
>(However, after further review, I may simply say neither parent
>nor child, merely links, which have names. See below.)
>There are some issues here that may be problematical.
>First off...the (P,C,iname) is unidirectonal, although it can be
>made bidirectional: (P,C,pname,cname), where C.pname is P, and
>P.cname is C. In short, *two* entries are created in the InodeNaming
>relation whenever an Inode is created, and both have to be named.
Yes. It gets more complex because there are sibling-sibling links. The
parent/sibling/child categories are necessary for the security scheme.
>Once one does that, the file system ceases to be a tree, and
>instead becomes a web -- not a problem, but can throw off some people.
>
>In fact, in a case like that, one might as well dispense with
>the notion of children and parents, and instead look at the issue
>from a graph-theoretic standpoint, where Inodes dot the data
>and Names dot the arcs, and a pathname is merely a series
>of traversal specifications -- arcs from the starting point.
You know, I never thought of naming directories instead of links.
The problem is that you might want some file to have multiple
different names, like glibc5.0 and glibc.
>(2) There are no directories as such. All nodes can have data,
> and arcs to and from them.
:-) It turns out that File is a subclass of Directory (the reverse
of Unix).
>>The only way to remove an object from the filesystem is to delete each
>>and every parent link to that object. Having a parent link to an object
>Partially correct; the only way to garbage-collect an object is to
>delete each and every cross-link thereto, making the object
>inaccessable.
If a container receives a delete, it checks whether it contains any
parent links (or is it child? I keep forgetting which is which). If
it does then that's it. If it doesn't then it marks itself as deleted
(updates the deleted timestamp) and sends a delete message to all its
links. It gets complicated because sibling links might not allow the
directory to delete them, in which case it sets the delete permission
on the other side of the sibling link (so that the other side's owner
/can/ delete the link) and then unsets every other permission, again
on the other side (so the other user can do nothing /but/ delete the
link). And then there's what the links do when they get a delete ...
I probably shouldn't mark the directory as deleted until it's finished
the operation. That way garbage collection is easy; something is garbage
if it is deleted (the delete time is greater than the last access time)
and it's been deleted before time X.
>>If the root has no parent, then it too is inaccessible. The root is not
>>a special case.
>
>The root has to be a special case; someone has to know where it is! :-)
>(One can simply have a convention that Inode 2 is the root; this means
>of course that an auxiliary data structure can be used to search for
>Inodes by internal Inode number.)
If someone knows where the root is, then it has a parent. <cackle>
>>Impossible since there would be no way to reliably figure out what
>>a user's domain is (in fact, if someone used encrypted directories
>>then the filenames would be scrambled).
>
>Hm...an interesting point, but this may break bidirectionality.
>Is an arc of my graph one of your "portals"? If so, that would
>make sense; a Portal (an Arc) would be nameable (though
>it's far from clear that the name must be unique on both sides).
>
>Does a Portal have a unique name across both its nodes, or not?
>An interesting question, methinks....
A portal subclass of Link, same as HardLink. The difference is that
portals are across filesystems, cycles don't matter, they may or may
not count towards the parentLinkCount (ie, if they count then you
could get an account on a machine, then cut yourself off from the
syadmin by disconnecting the remote filesystem), and they're built
using hard links.
An InsidePortal (that goes /into/ another filesystems) has the pathname
of the process' stdin) and the pathname within that filesystem of the
link to which it must forward any message it gets. The link it hands
off any messages to will be in a directory called /portals because of
the way portals are built. An OutsidePortal (that goes /out/ of an FS)
sends a message to the filesystem (which is just a process inside of
some other filesystem) to do I/O on its behalf.
>>a subclass File called EncryptedFile though I'm going to have to
>>think a lot about adding classes at runtime.
>
>You'd have to; how else are you going to install applications?
>(Besides, CORBA might be able to do most of the heavy lifting here.)
I don't think it can. Classes are not something an application
installs because it needs them at runtime, it's something a user
installs because he wants to make use of that class. I should
distinguish between classes local to an application and classes
an application installs in its environment.
The problem with being able to add classes is that you can't allow
anyone to add a method which does "Smalltalk destroyYourself".
>>What if you want to share the file with Alice, Carrie, Denise, Eric,
>>Gerhard, Irene, Julie, Louise, Marc and Veronica but you only want to
>>give Irene and Julie read access, and Gerhard can't get an account on
>>your system because your sysadmin hates you ...
>
>How does one share a file without read access? What a ridiculous concept,
I meant that you wanted to give everyone /else/ read and write access but
Irene and Julie read access only.
>unless the file is a pure executable. :-)
In my design, there's no execute permission; it simply doesn't make
sense in 90% of filesystems. OTOH, you have read, overwrite, delete,
grow, shrink and mutate. Mutate allows you to create a parent-child
link to that directory. For directories, grow permission is needed
to add /any/ link to that directory (similar for shrink-delete). For
files, grow allows you to append/prepend/insert data and you can
always create a new child-parent or sibling-sibling link to that
file (the permissions are duplicated into the new link so it's just
for convenience).
Permissions are stored in links so that different people (who may
only have access to different links) get different permissions.
There's also a concept of dominance wherein a container dominates
another container iff there is an unbroken parent-child path from
itself to that second container. Dominance overrides permissions
so you can do anything you want with files or directories you
dominate. A user's domain is what they dominate. It's actually
neat to implement; you send a request to a link telling them you
dominate them and if they're a child they'll do what you say but
if they're a parent they won't be intimidated ...
You know, I should just use BDSM language. <whip> <whip>
>If Alice, Carrie, Denise, Eric, Gerhard, Irene, Julie, Louise, Marc, and
>Veronica are part of one group (call it ExecutableGroup),
>and Irene and Julie part of another (call it ReadableGroup), then
>it's simple.
Sure. Who gets to create groups?
>True. I prefer a grouping structure similar to that used in
>modern Unices; the concept I have is an extension thereto,
>which allows all groups to have a representation within the
>File system (i.e., everything's visible in a uniform representation).
That's a problem; a representation is not enough. Far better if the
groups are embedded within the filesystem itself. In my design, the
filesystem doesn't have any concept of groups; groups are something
that emerges from the topology of the filesystem; if you and I both
have a parent-child link to the same directory then we're a group
(and owners of that group to boot).
>>You type them by class and this class means something different from
>>filesystem to filesystem.
>
>Oh lovely. Goodbye shareware! Goodbye datasharing! Goodbye
>installing software, period!
>Goodbye computer viruses!
>
>Say...that might be a *good* thing... :-)
>
>(Or did you have in mind base classes that would be similar within
>each system, and extension classes which differ from system to system?)
Every component in my design is a filesystem so in the LoggingFS,
a File is a segment. In the virtualFS, a File is pretty much your
everyday file. In the networkFS, a File is some kind of network
connection. In the windowingFS, a File (or more likely the Bitmap
subclass of File) is a window. Every module has its own definition
for what a file means.
------------------------------
Date: Thu, 03 Jun 1999 06:25:24 -0400
From: Jon Edwards <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: RH 6.0 and .bash_profile
Johan Kullstam wrote:
> Jon Edwards <[EMAIL PROTECTED]> writes:
>
> > I've loaded RedHat 6.0 and GNOME for the desktop with bash as
> > the SHELL. It appears as if .bash_profile simply doesn't execute.
> > Any ideas why not ? I there some file I need to modify to allow the
> > profile to execute.
>
> put interactive PS1, aliases and commands in .bashrc. all your shells
> will read this file. use the -norc flag in scripts to suppress
> reading .bashrc.
>
> put all login time stuff in .bash_profile. this means environment
> variables and things you want to start once per machine.
>
> > One of the things I'm trying to add is set -o vi If I add this to
> > .bashrc then I can't even issue set -o vi from an xTerm and have it
> > work.
>
> this i don't know about. i use emacs.
>
I must be an odd ball. I still can't get set -o vi to work unless I do not
put it in .bashrc and execute
it from the xterm. I can "live" with this, but I'm more concererned about
wanting to add more things
to the profile and have them execute. I know the bashrc is running as my
alias are showing up.
Should I try posting under a different news group ??
TIA !!
Jon
>
> --
> J o h a n K u l l s t a m
> [[EMAIL PROTECTED]]
> Don't Fear the Penguin!
------------------------------
From: Azfar Kazmi <[EMAIL PROTECTED]>
Subject: Re: Decent Partition Sizes??
Date: Wed, 02 Jun 1999 05:13:57 GMT
Reply-To: [EMAIL PROTECTED]
In article <[EMAIL PROTECTED]>,
Robert Heller <[EMAIL PROTECTED]> wrote:
> W32 (FAT16) 500meg /Win32 /dev/sda1
> 64meg / /dev/sda2
Wow! Isn't 64meg insufficient for /? / will have directories like /dev,
/bin, /sbin, /proc, /tmp, etc.
> 64-128meg swap /dev/sda3
Thats good. I have read that usually swap should be twice of physical
memory.
> 3.672-3.608gig *extended* /dev/sda4
Okay, I probably [or don't] understand the type extended. But what will
be its mouting point? Is the term 'extended' synonymous to what is used
in MS-DOS' fdisk? What is the use of it?
> 64 /var /dev/sda5\
With huge cache files of squid, I believe I have to have a much larger
/var!
> 1.5gig /usr /dev/sda6 | (In
/dev/sda4)
Probably, that's more than enough for /usr.
> whatever is left for /home /dev/sda7/
Okay.
> *If* you want to run your own news server, you'll want to get a
> *separate* disk for /var/spool/news.
No. No separate news server.
Thanks.
--
Azfar
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
------------------------------
From: Stephan Houben <[EMAIL PROTECTED]>
Crossposted-To: alt.os.linux.caldera,alt.os.linux,comp.os.linux.advocacy
Subject: Re: Does Java run well on Linux?
Date: 03 Jun 1999 13:33:11 +0200
[EMAIL PROTECTED] writes:
> Could you share your experience with me?
>
> Java programs dont run well on my windows 95 PC, they are slow and cause
> my system to hang. My PC has a 486 CPU w/ 64 MB RAM.
>
> I would like to learn Java programming. On this old
> 486 PC, i am wodering if it would be better to use Linux.
I've done some Java development on my Pentium with 32 MB.
First I used Win95, then I upgraded to Linux. While many things
(esp. LaTeX) went a lot faster, I didn't notice any speedup in
my Java stuff. They never crash my system; that's indeed an advantage.
If you don't need to do GUI programming, you could consider ditching
X, which will make more resources available for the Java stuff.
You could also consider using a language with a more modest footprint,
e.g. python. (Python can also be compiled into Java bytecode, IIRC).
Greetings,
Stephan
------------------------------
From: Azfar Kazmi <[EMAIL PROTECTED]>
Subject: Re: Decent Partition Sizes??
Date: Wed, 02 Jun 1999 05:21:03 GMT
Reply-To: [EMAIL PROTECTED]
It's good to know that I can change [increase/decrease] filesystem size
later. How can I do that? On another box, where I have only two
filesystems: / [500meg] and /home [500meg]. Now after Rehdat setup, / is
almost full and /home is almost empty. I wish to increase / and reduce
/home. [Not much users will be created on this box so /home will remain
unused mostly.]
--
Azfar
In article <[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Ian Briggs) wrote:
> It all depends on how you use your system, but FWIW these are the
sizes
> I'm currently using on a 3.1GB drive:
> swap 128MB
> / 400MB 9% full
> /usr 1500MB 60% full (with /opt linked to /usr/opt)
> /var 300MB 15% full (with /tmp linked to /var/tmp)
> /home 740MB 19% full
> so I'll probably end up reducing / and /var to 100MB each, and
increasing
> /usr to 2GB.
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
------------------------------
From: [EMAIL PROTECTED]
Subject: Kernel Hacking
Date: Tue, 01 Jun 1999 23:21:40 GMT
Hey to all you Linux gurus. It seems that I am having a bit of trouble
modifying my kernel.
I need to configure my kernel (the one that came with Linux 4.2) to my
sound card. I know how to do this. Unfortunately, the computer doesn't
want to recognize the "make config" or "make xconfig" commands.
Normally, I might download another copy of the kernel from the web,
unpack the source, and try again. However, being an AOL member, I can't
get access to the internet through Linux. And, Windows assimilates all
files that enters its filesystem. As soon as I download a file into
Windows, Win95 catalogs it. Assimilates it. After copying it to the
Linux partition, TAR won't recognize assimilated files, and RPMs become
useless as well. This is not an option.
BASH is properly installed. I'm in the right directory, /usr/src/linux.
How can I get Linux to recognize the "make xconfig" command?
Thanks for all your help. Please respond via email, as I rarely check
replies to my posted messages. I'm more likely to read a solution if it
is in my mailbox.
Thanks.
Steve.
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
------------------------------
From: "Jan Rockstedt" <[EMAIL PROTECTED]>
Subject: scsi error on hd
Date: Thu, 3 Jun 1999 12:56:11 +0200
Hi !
I have a test linux pc.
Redhat 6.0 and one scsi hardisk.
Sometime i get scsi error on some sectors.
Is there any apps that fix this problem (scandisk in pc or ... ) ?
The fsck do not report any problem.
Regards Jan Rockstedt
------------------------------
From: [EMAIL PROTECTED] (Carl Fink)
Crossposted-To: comp.os.linux.setup
Subject: Re: INN news server
Date: 3 Jun 1999 07:08:12 GMT
Reply-To: [EMAIL PROTECTED]
On Wed, 02 Jun 1999 01:25:24 -0400 Patrick Lanphier <[EMAIL PROTECTED]> wrote:
>Where do I find information on how to configure the INN news server? I'll I would
>like to do is
>setup a couple of private news groups.
Well, the obvious:
man innd
man inncheck
man expire.ctl
man inn.conf
etc.
My Debian inn package came with lots of documentation in /usr/doc/inn.
I don't know where you're getting inn, of course. There's an INN FAQ
posted regularly to news.software.nntp, or you could check rtfm.mit.edu.
And any specific questions should really go to news.software.nntp, not
one of the comp.os.linux* groups, unless they're Linux or distribution
specific.
--
Carl Fink [EMAIL PROTECTED]
"This fool wishes to reverse the entire science of astronomy."
-Martin Luther on Copernicus' theory that the Earth orbits the sun
------------------------------
From: [EMAIL PROTECTED] (Carl Fink)
Subject: Re: Is This Illegal?
Date: 3 Jun 1999 07:12:59 GMT
Reply-To: [EMAIL PROTECTED]
On 2 Jun 1999 21:23:53 GMT K Lee <[EMAIL PROTECTED]> wrote:
>
>Just so my post was not misconstrued as I am intending, although I could
>see how it may have been, to burn distro CDs to distribute for profit, I
>refer to you the links below:
Just so you know: several companies are already doing this, and
you'll have a hard time competing with prices like LSL's or
Cheap*Bytes's, which range as low as $1/disk.
See www.lsl.com or www.cheapbytes.com.
--
Carl Fink [EMAIL PROTECTED]
"This fool wishes to reverse the entire science of astronomy."
-Martin Luther on Copernicus' theory that the Earth orbits the sun
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.misc) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Misc Digest
******************************
