If I could re-write Linux (NewsForge)
Linux is a pure 32-bit operating system written from scratch for 32-bit processors. It doesn't suffer from any 16-bit baggage code. Now Linux is being ported to various 64-bit processors. It will be a while before all the code is compiled and optimised to take advantage of 64-bit platforms."
Posted Oct 24, 2003 18:36 UTC (Fri)
by rfunk (subscriber, #4054)
[Link]
Posted Oct 24, 2003 19:28 UTC (Fri)
by einstein (subscriber, #2052)
[Link] (2 responses)
Posted Oct 24, 2003 21:27 UTC (Fri)
by dmantione (guest, #4640)
[Link] (1 responses)
Posted Oct 27, 2003 4:33 UTC (Mon)
by gdt (subscriber, #6284)
[Link]
Yep. Which computer has been so in demand that the maker can't keep up? The Apple iPod. The beautiful thing about combining Linux with 32b handheld hardware is that it allows the merge of a real OS with a real form factor. PDAs get the same self-repairing filesytem as used on huge clusters, ADSL modems run the same DHCP server used by big enterprises. To my mind, the push of form factor requirements back onto the OS is where the next set of interesting changes to operating systems is going to come from. Linux isn't particularly well set up for this: there are still too many assumptions that the console is secure (as opposed by being poked by a three year-old who likes the movement on the screen), even the lack of willingness to integrate a crash dump shows a strong assumption that development is done on 80x24 screens.
Posted Oct 24, 2003 19:36 UTC (Fri)
by torsten (guest, #4137)
[Link] (6 responses)
"NGOS would also follow a modular design but would do away with the shell prompt altogether. Its GUI would talk directly to the operating system."
That's about as far as I read. I was confused initially, trying to figure out what is wrong with him. Now I know. He's would rather be using Windows.
Posted Oct 24, 2003 19:56 UTC (Fri)
by proski (subscriber, #104)
[Link] (5 responses)
The only reasonable proposition is to have an access counter for files. The rest is either irrelevant to the kernel (i.e. GUI stuff) or is based on the assumption that dropping support for something (NFS, text mode, legacy filesystems) is going to make the OS better.
As for the suggestion to make the new OS more secure than OpenBSD, I can only say "good luck".
Posted Oct 24, 2003 21:28 UTC (Fri)
by TheOneKEA (guest, #615)
[Link] (4 responses)
Posted Oct 25, 2003 1:23 UTC (Sat)
by tracyreed (guest, #16253)
[Link] (3 responses)
telnet selinux.copilotconsulting.com I invite you to try to undermine the system or get into data you shouldn't be able to access such as /etc/shadow.
Posted Oct 25, 2003 1:51 UTC (Sat)
by StevenCole (guest, #3068)
[Link] (2 responses)
Posted Oct 25, 2003 10:02 UTC (Sat)
by TheOneKEA (guest, #615)
[Link]
I played with SELinux on 2.4 a long time ago, but it was too difficult to eliminate the hundreds of permissive policy violations that kept appearing in the dmesg. I wonder if anyone has written SELinux policies for various distros; i.e. one for Red Hat, one for Mandrake, etc. That would make it a LOT easier to set up.
Posted Oct 25, 2003 10:33 UTC (Sat)
by error27 (subscriber, #8346)
[Link]
Posted Oct 24, 2003 19:57 UTC (Fri)
by ccchips (subscriber, #3222)
[Link]
Posted Oct 24, 2003 19:57 UTC (Fri)
by stumbles (guest, #8796)
[Link]
Posted Oct 24, 2003 21:13 UTC (Fri)
by TheOneKEA (guest, #615)
[Link]
OTOH, his comments about security did have some merit, but again, some research would have shown that LSM, SELinux and friends are filling that gap quite nicely.....
Posted Oct 24, 2003 21:23 UTC (Fri)
by dmantione (guest, #4640)
[Link]
Posted Oct 25, 2003 1:23 UTC (Sat)
by StevenCole (guest, #3068)
[Link] (5 responses)
Hmmm. He's on to something. Maybe we should look
to societal trends to vision the future of our NG OS.
My local public library no longer has a sign labeled "Library".
Instead, a silhouette of a person reading a book indicates that
the library is near. That's so non-literate people looking
for the library will know they're getting warmer.
Perhaps Western civilization has been on the wrong track for
the last 3,000 years since the invention of the alphabet.
Look at all the problems it's caused. Last time I was in
Russia looking for a restaurant, all I saw was signs
with "PECTOPAH". Now, if only they had a picture of someone
eating a good old hamburger. What could be more universal
than that? So, the NG OS should do away with characters,
and therefore the keyboard as we know it could be trashed.
A nice, simple replacement with perhaps as few as 500 symbols
could replace it. Without a character set, word processors
and editors would be a thing of the past, along with the
usual vi vs. emacs debates.
He really knows his stuff when it comes to the benefits
of a microkernel. If only Linus had listened to Professor T.
And he's right about the increased ease of porting to
other platforms. A quick check in the arch directory of
2.6.0-test8 shows a paltry 20 entries. Since the
superiority of micro over monolithic is so obvious, the
only remaining detail is how to optimize the communication
between the various components. Some have suggested XML,
but I feel this is far too low level. If we put natural
language processors in each kernel component, the communication
could be done in a natural, and hence better, way. Now,
this could be the source of some slight loss in
speed, so the kernel components could have some embedded AI which
would learn to use slang to speed things up a bit. This
could be extended to user space, so that the NG Mozilla,
instead of doing a malloc for another gigabyte of memory,
could just do a system call which said (literally) "Gimme
my usual". And the kernel, knowing it came from NG Mozilla,
would know what to do.
And speaking of knowing what to do, the little guy hiding
under my bridge is hungry. Time to feed him.
Posted Oct 25, 2003 19:42 UTC (Sat)
by dlang (guest, #313)
[Link] (4 responses)
Posted Oct 26, 2003 1:38 UTC (Sun)
by StevenCole (guest, #3068)
[Link]
When I've traveled to Europe, I've always made a point of
learning a very basic vocabulary, both written and spoken,
not wanting to repeat the mistakes of my youth, when I
believed that Peligro meant "Free Swimming", since
all the good swimming holes were so labeled.
What I detest is the gratuitous use of picture signs when
a simple word will do. Sure, my above example seems to
counter that, but simply adding "Danger" to the sign,
especially since it was right here in the USA, doesn't
seem to be asking too much.
Anyway, those were the most ridiculous suggestions I
could think of at the time. Eliminating an alphabet
based character set might provide the coup de grace
to the hated command line. Or maybe not.
I forgot about suggesting adding
floating point to the kernel. And NG hardware should only
respond to suggestions, not instructions, which are an
artifact of older pre-choice days.
Posted Oct 27, 2003 0:30 UTC (Mon)
by komarek (guest, #7295)
[Link] (2 responses)
At any rate, I expect that people type "rm -rf" regardless of their locale, and that is beautiful. =-) -Paul
Posted Oct 28, 2003 12:01 UTC (Tue)
by morhippo (guest, #334)
[Link]
Posted Nov 7, 2003 13:47 UTC (Fri)
by forthy (guest, #1525)
[Link]
The chinese characters are a mix of pictures (or pictures that were recognisable
but
aren't now due to the way they are written now - the signs of sun and moon pretty much
look like sun and moon in ancient chinese, but today, you can only recognise that the
basic strokes are alike), and phonetic components. That's since most glyphs are
combinations of several (smaller) glyphs. You can have composed glyphs that mean
something (e.g. pig under roof means "home", or woman under roof means "peace"), or
you have composed glyphs that indicate how it's spoken (e.g. yuan - the money unit
sign - in a rectangle means "garden", also pronounced "yuan"; but yuan - together with
a man - is part of the sign for "bank"). Almost all glyphs have at least one component
that indicates what it should mean, and 80% of the glyphs contain an indication of how
to pronounce them (in 2200 year old chinese from Xian, not necessarily lokal dialects or
today's chinese). At least when you know what the component glyphs are supposed to
stand for as picture (which isn't very recognisable from modern Chinese), or are
supposed to be pronounced. Ancient (old kingdom) Egypt Hieroglyphs are different stuff - they are a mixture of
consonants, double-consonants, and often used words (i.e. mostly phonetic, only very
frequently used words non-phonetic). Unfortunately, the phonetic part is only
guesswork. The glyphs for frequently used words are recognisable pictures (not
always). A chinese monk once complained that all those long-noses took pictures of his
temple,
despite it was written in bold words "no photographs" (he wanted to sell postcard). In
Chinese words, of course. He complained "can't they read, er?" A camera with red
circled stroke through can be understood - but basically, do you expect a camera to
look like that in 2200 years, or people to understand that a red circle with a stroke
through from lower left to upper right means "not"?
Posted Oct 27, 2003 4:55 UTC (Mon)
by jimi (guest, #6655)
[Link]
As the comments on Newsforge point out, this is a horribly uninformed If I could re-write Linux (NewsForge)
article.
Oh geez, we had 64-bit Linux on Dec Alpha back in what, 1995?
If I could re-write Linux (NewsForge)
Absolutely! But also: If I could re-write Linux (NewsForge)
"Small computers are the future of the world", Chuck Peddle, inventor of the 6502
processor, 1975.
If you throw everything towards infinte memory and filesystem sizes, you throw away
one of Linux strong markets, embedded.
Article assumes computers won't change
Article sounds like a Microsoft roadmap.
Yes, that was lame. The author confuses shell and text mode support. Text mode may become obsolete some day (you can disable it in the kernel already), but shells are here to stay.Article sounds like a Microsoft roadmap.
Things like grsecurity, LIDS and SELinux. Has anyone tried to hAx0r a box running those and then tried to hAx0r a box running NetBSD?
What about the LSM-based stuff?
LSM based stuff rocks. I use SE Linux and find it absolutely essential to having a secure machine. If a system does not have a means of handling the inevitable buffer overflow in a program running as root and listening on a socket it's not secure enough for me. That's why I use SE Linux. I have a demo SE Linux system on the net for you to play with. I'll even make it easy and give you the root password:What about the LSM-based stuff?
user: root
pass: root
Yep, that seems pretty cool.
What about the LSM-based stuff?
root@selinux:~# df -T
Filesystem Type 1K-blocks Used Available Use% Mounted on
/dev/hda1 reiserfs 2104408 744828 1359580 36% /
/dev/hda3 reiserfs 35920236 32860 35887376 1% /home
root@selinux:~# umount /dev/hda3
umount: /dev/hda3: not mounted
umount: /home: must be superuser to umount
root@selinux:~# whoami
root
root@selinux:~# cat /etc/shadow
cat: /etc/shadow: Permission denied
root@selinux:~# ls -l /etc/shadow
ls: /etc/shadow: Permission denied
root@selinux:~# chmod u+r /etc/shadow
chmod: failed to get attributes of `/etc/shadow': Permission denied
Hehehehe :DWhat about the LSM-based stuff?
Hm... When I type it, it's a little bit different.
What about the LSM-based stuff?
root@selinux:/# df
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/hda1 2104408 2103504 904 100% /
/dev/hda3 35920236 32860 35887376 1% /home
root@selinux:/#
But it's pretty cool none the less. :)
...every day would be the first day of spring....
If I could re-write Linux (NewsForge)
I think he just wanted a reason to use the acronym NGOS.
If I could re-write Linux (NewsForge)
Reading through that article showed the incredible lack of research he has done, combined with some rather poor ideas. Most of the limitations and issues he cites are either nonexistent, ignored or just plain wrong.IMO, this guy doesn't know what he's talking about
Hmmm. The idiot complains about the (minor) performance issues with kernel modules, If I could re-write Linux (NewsForge)
instead he wants a microkernel. I would be incredibly suprised if all those userspace
drivers would be faster than the Linux kernel modules.
Astonishing. Simply astonishing.
If wishes were fishes...
you forgot to point out that the Chinese already tried the picture based writing as opposed to an alphabet
If wishes were fishes...
Don't know much about Chinese, but
I'm somewhat aware of Japanese Kanji (derived from Chinese)
and the fact that the Japanese recognized the value of
an alphabet 1200 years ago and so developed Hirigana
and Katakana, although their implementation details and usage
are beyond the scope of any short discussion.
If wishes were fishes...
I'm not clear on your post -- the Chinese tried picture writing at some time in the past? They certainly are not drawing pictures now. There's no alphabet (there is a phonetic alphabet, though), but the characters are *not* pictures. The only widely-used picture writing I know of are the Heiroglyphics of Ancient Egypt.If wishes were fishes...
are not pictures. It is a syllable alphabet and too a very large degree purely phonetic.
Hieroglyphes ...
If wishes were fishes...
Many of these suggested changes are horrible ideas. Why didn't they just title this article, "I Want A Windows Version Of Linux"? Microkernel vs. monolithic kernel designs have been hashed over before with great advantage going to either design. Doing away with a text mode interface is awful as it assumes you have graphics-ready hardware, thereby eliminating flexibility. And have we forgotten that scalability is two directional, up and down? And should a filesystem be optimized for user searches? Do we spend most of our time searching around for files? The article suggests that everything should be an object, is the author aware that in unix-like OSes everything is an object: a file?
Newsforge: let's make a windows version of linux