response 28 of 257:

Jan 31 15:15 UTC 1999

        Is this "Linux Central" you mention webwise? I was looking for a
        CD of the SuSE stuff, but cheapbytes sells only a multi-cd package
        & book combo.. I'd like to get a variety of linux/FreeBSD CD's
        so as to be able to switch over - over even install for others -
        and have much of what I already search the net for around..

        On another note, installing the new 2.2.1 kernel is becoming a
        real challenge. it appears that, not only does it use a different
        module-loading scheme (kmod instead of kerneld), but.. My system
        seems to NEVER have been using modules right under my ol' reliable
        2.0.35 kernel.. Lotsa' munged-up stuff, and I've been debugging
        and cleaning up the /etc/rc.d/* stuff as I go along.

        Oh, and I _STILL_ can't figure out wtf generates the
        /boot/module-info files.. Plus, I get exciting messages like this
        one (in /var/log/messages):

Jan 31 09:50:22 localhost modprobe: no dependency information for module:
"/dev/modem"

        Further, I've seen warnings that the 2.2.1 stuff - or the modutils
        - can break your 2.0.x stuff.. Could be the case in some places,
        here.. Still, I'll be damned if I can find a decent doc on the
        "Use & Feeding of /etc/conf.modules" - it's like everyone should
        already "know" it.

        Additionally, by the time you get all the libs and packages
        upgraded in support of the 2.2.1 kernel, you'll suddenly discover
        that some of yer stuff is now broken good & proper.. Xmahjongg
        went first; then pico; then pico was reborn; then glint (and
        friends) refused to play; and lately, after upgrading gpm, 'mc'
        has decided to scream and die. Oh, and you ain't seen fun until
        two distinct packages decide to argue over whom is going to update
        the same two programs - THAT was fun..

response 30 of 257:

Feb 2 03:36 UTC 1999

MODVERSIONS is an option in the kernel config.  Simply answering 'yes' to
the 'Config Kernel Modversions?' question as you do the 'make config'
would accomplish the same thing as commenting out the test as you did.

response 32 of 257:

Feb 2 20:58 UTC 1999

Re the above discussions on partitioning...my own practice is that if I'm
setting up a server, I seperate out /usr.  This does two things; it makes
backups easier to handle for me, and if /usr gets corrupted somehow the
machine will still boot, since the damage will be contained away from /. 
It's a good idea to have a seperate /var in this situation, too, for the
above reasons, and also so a full mail spool won't cripple the whole system. 
/home is another natural to split off, but on my system it's part of /usr. 
This is partially because of how I set up the machine.  I'm using old, small
disks; /, /var, and swap are partitions on the first drive, which is 130
megs.  /usr is the entire second drive, which is 450 megs.  Obviously you
have a lot of flexibility here if you know what you're doing, but it takes a
while to get a feel for how big each partition should be.

For my desktop machine, where I'll be the only user, these aren't issues, so I
generally make one big filesystem, so I don't run out of space as easily.  I
gig should be enough to experiment with Linux.  I used to use it extensively
on a 400 meg partition.  (Of course, on my current machine I have 1 gig for
Windows and 5 for Linux.) ;)

The swap space issue is a little less clear-cut than most people make it. 
2x the physical RAM is an okay starting point, but use your judgement.  If
you have an 8 meg machine, you may need more memory than this, so you might
want to make a 20 meg swap partition.  If you have 64 megs of RAM, you're
probably not ever going to need 128 megs of swap, unless you're doing
something amazingly heavy, like editing huge image files.  My 64 meg machine
has 64 megs of swap, and it rarely uses any of it.  (Let me tell you, a
machine that doesn't have to swap is *ungodly* fast.  If you're swapping a
lot, upgrade your RAM before you even think about a processor upgrade.)

response 34 of 257:

Feb 3 16:03 UTC 1999

        Hmm, check yer kernel's out, and distributions, too..

        From what I can see, they typically bypass the bios, and I've had
        to save space my removing all SORTS of different arch/
        directories.

response 36 of 257:

Feb 4 11:13 UTC 1999

At work, I run openbsd on a IIci, with 8M ram & 258 M disk.  I'd
consider that more or less a minimal system for 4.4bsd.  The IIci has a
68030.  The 68020 is similar, but lacks an MMU.  You need the MMU to run
unix.  The install setup with openbsd (and friends) uses a ram disk; so
8 M of ram is pretty much the minimun.  (It *might* run in 5 M; but
definitely won't work in 4 M.  I think the IIci takes memory in 4 M
increments.) The basic system seems to be about 90 M, including the
compiler.  There is some fat here; perhaps this could be trimmed down.
But it's cheaper timewise to just get a larger disk.  You could also try
this on a q700, which is basically a faster IIci.  In theory, X would
run on the IIci, but it would be really slow, and nobody seems to have
bothered putting work into frame buffer support for openbsd.  Besides
the macintosh line, you should also look at getting a cheap 386 system.
It's a lot easier to find peripherals for the isa bus.

The original Mac II's were shipped without MMU's, but could have one
plugged in after-market; so some will have that and some won't.  Apple
never sold any 68010 systems; but, if you ran across such a machine,
again, you'd need an MMU, support in the kernel for the MMU on the
machine, & it will be very slow by modern modern standards.  (The 68020
is, roughly, a 2mips processor; the 68010 is about a .5 mips processr,
or one quarter the speed.) The SUN-2 design was based on the 68010.  The
kernel that Sun sold on this machine was basically 4.2bsd.  The 68010
can address a maximum of 16M of ram; however, with the memory chips of
the time, 2M systems were most common.

The 68000 can't restart instructions after a page fault, so can't
support demand paging.  Systems based on the 68000 included the altos
68000 (that m-net used to run on), and the sun-1 design.  The altos ran
system III; the sun-1 ran unisoft unix, basically version 7 unix.
Neither of these versions of unix supported networking.  Networking came
in sometime after demand paging, so it would be very hard to find tcp/ip
in any version of unix for the 68000.  When the 68000 was popular,
networking most often meant uucp.

None of the 68000 macintoshes had MMU's, so getting a "real" version of
Unix to run is most unlikely.  Getting the "fork" primitive in Unix to
work basically requires having some form of memory mapping hardware.
Either that, or *really* fast swap I/O (which is what they actually did
do on one of the cray's.)

(The reason I used the IIci at work was not because of its outstanding
performance, but because it turned out to be easier to scrounge the
hardware and several ethernet adapters to put together a really cheap
router based on the IIci, than to put together the same thing with the
386.  I think this is because 386's were worth more $ at property
disposition.)

response 38 of 257:

Feb 5 21:10 UTC 1999

>  (1)  Why was df showing that /var was at 109% capacity?
>  Any ideas?

In BSD (and most UNIXes, in fact) there's a certain amount of space on
the filesystem that's reserved for the superuser.  This is a sort of
'cushion' to help prevent users from taking down the system by filling up a
filesystem.  What you were seeing is a truely full filesystem: all 100% of
the user space was full, plus the 9% of reserved space.  The amount of
reserved space can often be adjusted when the filesystem is created, but
generally there's little reason to do this.

response 40 of 257:

Feb 5 21:27 UTC 1999

I have re-installed Win NT several times without the MBR (containing LILO)
being touched.

response 42 of 257:

Feb 6 05:49 UTC 1999

        Had Slackware installed, once-upon-a-time.. 

        Used LILO.

        Upgraded the DOS/DOZE partition to win95..


        *BZZZZZZZZZZZZZZZZZZZZZZZZTTTTT!!!!!!!!!!!*

        Slackware, Lilo & Linux "Go bye-bye".


        Lilo writes a bootstrap record to MBR that routes to the lilo
        & "which one" compressed kernel.

        M$ products couldn't care less wtf you've installed - they will
        mangle whateverthehell they want to..

        Moral of the story:

                Create a Zdisk.. The result can find root and all else
                after the MBR has been screwed, blued & tatooed.

        Empirical Evidence that the Zdisk doesn't care to work following
        creation & reboot: I couldn't boot from the floppy.

        Empirical Evidence that the Zdisk DOES work after you give up in
        frustration: I got peeved & went to bed - PowerOn the next day
        booted precisely the kernel I'd compiled (2.1.1) from the
        diskette.

response 44 of 257:

Feb 6 22:50 UTC 1999

Ouch.  That's a bad one.  I'm glad I've never used Plus.

response 46 of 257:

Feb 8 21:19 UTC 1999

Yes, but it requires a portion in the MBR.  The good thing about LILO is 
you can install it exclusively in the boot sector of your linux root 
partition.  Then, when Windows changes itself automatically to be the 
boot OS, all you need to do is change the active partition, and you are 
back to linux.  You can't do that with FreeBSD.  FreeBSD and Windows 9x, 
in my experience, don't mix as well as Linux and Windows 9x.

response 48 of 257:

Feb 11 14:03 UTC 1999

        rrrhh?

        You done lost me with "superblock"..

        Best solution that I've had is to start with a small DOS partition
        and then create the linux parts - with /boot being one of 'em..

        Then, you keep the kernel/boot stuff in /boot and let lilo loose
        on the MBR..

        If 'Doze whacks the MBR, the only solution I've seen is to have
        the bootable "zdisk" around.. I'm fairly sure I could get lilo
        to retake the high-ground of the MBR, but <shrug> Not being in the
        least concerned with win<pick-a-number>, I doubt it will ever be
        an issue again..

        Actually, my concerns anymore deal with RH over Debian/TL and
        FreeBSD.. And, THAT particular donneybrook is going to get intense
        this weekend.

response 50 of 257:

Feb 11 16:47 UTC 1999

        Crap... That is Not A Good Thing (tm)

        You realize, of course, that I did NOT want to hear this...
        as I am currently trying to decide twixt FreeBsd, RH 5.2, Debian
        and TL as my ONLY Operating System..

        Goddamnit... This Bodes Not Well at all.. *sigh*

        Well, maybe if I use FIPS and shrink down the DOS partition to the
        minimal for Dos 6.0... crapdoodle.. I wanna' be m$ free.

        I would point out, to those that ain't experienced it, that trying
        to get any sorta' older DOS installed from any of the win9x crap
        is a total wash.. And, for the love of Bog - make sure you have a 
        bootdisk with that old DOS that WORKS before suffering m$ any
        further..

response 52 of 257:

Feb 11 17:46 UTC 1999

        Yeah, I recall seeing it long ago and hearing of it..

        I have no clue how it is doing, let alone how it "plays well with
        others". Certainly win95/98 does NOT play well.