It is open source, except for a userspace device driver for the forwarding ASIC. Without the driver, everything works the same, you just don't get hardware accelerated forwarding, only the normal kernel softward forwarding.
The biggest difference vs EOS is that if you want to add a route to the routing table in EOS, you have to use sysdb-specific commands/APIs. With Cumulus Linux, you use "ip route add" or any other program that knows how to add routes to the Linux kernel using netlink or legacy methods. Same with ACLs, EOS has proprietary commands/APIs, Cumulus Linux uses iptables.
Also, A random Linux program will install and work fine on Cumulus Linux, whereas it usually takes a (small, but real) amount of work to make that happen on EOS. I've even installed and run Firefox from the Debian repo onto a switch, and it worked fine.
The big difference is that there is a hardware forwarding chip involved. A PC with 10G NICs is hard pressed to forward at 80 Gbit/sec, and draws a couple hundred watts. The 1U switches Dinesh is talking about can do 1.28 Tbit/sec with all features enabled, and draw around 100 watts.
Try dropping down one of the menus in windbg. Don't select one, just leave the menu up. 100% CPU.
It gets even better; windbg runs with elevated priority, so if you try this in a normal priority VNC server, it is time to call and have someone move the mouse for you on the physical machine.
"None of the venurabilities are relivant since none of the services they affect are turned on or allowed through the firewall."
You have an unpleasant surprise in store when the next virus (like nimda) comes right through your firewall in an email attachment and then starts exploiting your vulnerable "firewalled" systems right and left.
I've seen that exact scenario TWICE now.
Pride, the fall, etc, etc. be careful. its a dangerous internet out there, and firewalls are just one layer in effecitve security.
Its nice that you like P4, and it is excellent software, but it doesn't solve the problem here.
The problem is that neither P4 or CVS do binary diffing in their repository (IIRC, they both use RCS files on the backend, though P4 stores metadata in a transactional local DB).
Subversion can help here, since it stores binary file deltas where appropriate. Unfortunately, its just about to become "pre-alpha" so suggesting it for a production environment is premature at best.
Perhaps one of the other commerical source control systems supports binary diffs? Try google.
You've nailed it. Its odd your post is still hanging around at Score: 1 even though its the only one that correctly explains the glaring flaw in the benchmark.
ext2 does no journalling, so all 10 megs of your writes were buffered into ram. The benchmark then returned done, even thought little or no data had reached the disk yet.
The journalling filesystems do the same thing, with one major exception. The metadata changes MUST be written to the journal before the benchmark returns done. The data can stay in memory, since XFS and reiserfs only do metadata journalling.
You're comparing the time to write 10megs to RAM, read it back from ram and then remove it from ram against having to write 20000 files worth of metadata (10000 for the creates, another 10000 for the deletes) to disk, plus all the RAM writes/reads..
Maybe its my database background speaking, but non-normalized (for the uninitiated read: redundantly stored, more or less) data storage is bad news. If you have redundant data, the different copies WILL eventually become out of sync.
UNIX file (1) uses the contents to determine type. Any other form of meta-data is redundantly storing the type and the contents.
Its a pain when I download a wav file on my mac and the OS thinks Netscape should open it. The solution is to not store type metadata redundantly where it can become out of sync with the real type (determined by the actual comments).
Its by one of the RHAD labs kids. Its basically just a GUI around bohem's garbage collector in leak-detector mode.
Its not purify (it really aims for leak detection, not all the other errors purify finds), but the efence + memprof combination gets you about 85% of purify's functionality.
It seems to handle threaded apps reasonably well, and C++ doesn't faze it. The only down side is that its hard to get running on non-x86 platforms.
Their website sucks, but the software can transcode between MP3, AC3, WAV, and RealAudio. It can also transcode video, supporting OpenDiVX, MSMPEG4 (real DIVX), MPEG2, RealVideo and MJPEG. All without stupid windows codec hacks.
Its mainly designed to be a realtime encoding and streaming system, but it includes a utility that can be used like sox.
I'm not sure why this hasn't really made a splash in the free software world yet, since the codecs included in ffmpeg mean people can play DiVX videos without any windows files, and on other CPU architectures.
It describes the "ETag: " header, which is usually a md5 hash of the resource.
The client can then validate the resources in its cache by sending a request with a "If-None-Match: " header with the ETag associated with the copy in its cache.
The server will either respond "Not modified" in which case the client simply uses the version in its cache, or the server will resend the resource if the ETags don't match.
Since this patent was filed for in 1999, this is pretty clear prior art, in the most commonly used protocol on the largest network in the world. If the patent office can't locate prior art in incredibly obvious (obvious to anyone skilled in the art, that is) cases like this one, what hope do we have for them intelligently handling more subtle cases?
actually, on my powerbook, I map the "spare" enter key (just to the right of the space bar) to the right mouse button. When using the mouse, I place my thumb on the mouse button, my pointer finger on the trackpad, and my ring finger on the enter key. I find this arrangement even better than a standard two button trackpad, since I don't have to move any fingers to hit either button.
The only bummer is that I had to map F12 as the middle click, since the extra ALT button next to the spare enter key generates the same keycode (at the OF level) as the main ALT key. Fortunately pasting is not so common that its a huge problem, but it makes using xfig a PITA.
Anyone know of a way to hack OF to make it generate different keycodes for the left and right ALT buttons? I know the ADB emulation happens there, so I'd bet its possible.
Fading mouse trails are _NOT_ motion blur. I blame 3dfx and their "T-Buffer" bullshit for this misconception. Simply mixing each frame with faded versions of the last few frames is just plain ugly. It certainly isn't motion blur.
VSA100's motion blur isn't really motion blur. its just an accumulation buffer.
They just blend down the last n frames. if n = 4 you see 4 copies of the object, overlaid, each more transparent than the last. Actual motion blur is temporal antialiasing. The best way to do it is to draw each pixel n times, jittered a random amount backwards in time (a random amout of time, not a random amount of frames!).
They only hand out the awards for best support on BEA newsgroups. Posting this weblogic fanboy bullshit on slashdot isn't gonna win you the big prizes.
you can add a 'z' to the front of xvf and cvf if you want it to compress as it streams. may or may not be a good idea depending on how fast your CPUs are relative to your network.
Slashdot Mirror
# ip link
1: lo: mtu 16436 qdisc noqueue state UNKNOWN mode DEFAULT
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: eth0: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 1000
link/ether 70:72:cf:8c:23:4d brd ff:ff:ff:ff:ff:ff
3: swp1: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:4e brd ff:ff:ff:ff:ff:ff
4: swp2: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:4f brd ff:ff:ff:ff:ff:ff
5: swp3: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:50 brd ff:ff:ff:ff:ff:ff
6: swp4: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:51 brd ff:ff:ff:ff:ff:ff
7: swp5: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:52 brd ff:ff:ff:ff:ff:ff
8: swp6: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:53 brd ff:ff:ff:ff:ff:ff
9: swp7: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:54 brd ff:ff:ff:ff:ff:ff
10: swp8: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:55 brd ff:ff:ff:ff:ff:ff
11: swp9: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:56 brd ff:ff:ff:ff:ff:ff
12: swp10: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:57 brd ff:ff:ff:ff:ff:ff
13: swp11: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:58 brd ff:ff:ff:ff:ff:ff
14: swp12: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:59 brd ff:ff:ff:ff:ff:ff
15: swp13: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:5a brd ff:ff:ff:ff:ff:ff
16: swp14: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:5b brd ff:ff:ff:ff:ff:ff
17: swp15: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:5c brd ff:ff:ff:ff:ff:ff
18: swp16: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:5d brd ff:ff:ff:ff:ff:ff
19: swp17: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:5e brd ff:ff:ff:ff:ff:ff
20: swp18: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:5f brd ff:ff:ff:ff:ff:ff
21: swp19: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:60 brd ff:ff:ff:ff:ff:ff
22: swp20: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:61 brd ff:ff:ff:ff:ff:ff
23: swp21: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:62 brd ff:ff:ff:ff:ff:ff
24: swp22: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:63 brd ff:ff:ff:ff:ff:ff
25: swp23: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:64 brd ff:ff:ff:ff:ff:ff
26: swp24: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
link/ether 70:72:cf:8c:23:65 brd ff:ff:ff:ff:ff:ff
27: swp25: mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
It is open source, except for a userspace device driver for the forwarding ASIC. Without the driver, everything works the same, you just don't get hardware accelerated forwarding, only the normal kernel softward forwarding.
You can get the patches against Debian Wheezy here:
http://oss.cumulusnetworks.com/
The biggest difference vs EOS is that if you want to add a route to the routing table in EOS, you have to use sysdb-specific commands/APIs. With Cumulus Linux, you use "ip route add" or any other program that knows how to add routes to the Linux kernel using netlink or legacy methods. Same with ACLs, EOS has proprietary commands/APIs, Cumulus Linux uses iptables.
Also, A random Linux program will install and work fine on Cumulus Linux, whereas it usually takes a (small, but real) amount of work to make that happen on EOS. I've even installed and run Firefox from the Debian repo onto a switch, and it worked fine.
- nolan
CTO/Cofounder, Cumulus Networks
The big difference is that there is a hardware forwarding chip involved. A PC with 10G NICs is hard pressed to forward at 80 Gbit/sec, and draws a couple hundred watts. The 1U switches Dinesh is talking about can do 1.28 Tbit/sec with all features enabled, and draw around 100 watts.
- nolan
CTO/Cofounder, Cumulus Networks
Try dropping down one of the menus in windbg. Don't select one, just leave the menu up. 100% CPU.
It gets even better; windbg runs with elevated priority, so if you try this in a normal priority VNC server, it is time to call and have someone move the mouse for you on the physical machine.
"None of the venurabilities are relivant since none of the services they affect are turned on or allowed through the firewall."
You have an unpleasant surprise in store when the next virus (like nimda) comes right through your firewall in an email attachment and then starts exploiting your vulnerable "firewalled" systems right and left.
I've seen that exact scenario TWICE now.
Pride, the fall, etc, etc. be careful. its a dangerous internet out there, and firewalls are just one layer in effecitve security.
Its nice that you like P4, and it is excellent software, but it doesn't solve the problem here.
The problem is that neither P4 or CVS do binary diffing in their repository (IIRC, they both use RCS files on the backend, though P4 stores metadata in a transactional local DB).
Subversion can help here, since it stores binary file deltas where appropriate. Unfortunately, its just about to become "pre-alpha" so suggesting it for a production environment is premature at best.
Perhaps one of the other commerical source control systems supports binary diffs? Try google.
Intel's compiler boosts AMD Athlons too.
AMD uses (or at least, used to use, I haven't checked lately) Intel's compilers for their SPEC runs.
Intel's compiler is the best available for CPUs that implement the x86 ISA. Transmeta implements that ISA, so why does this news surprise people?
You can use newlines in a filename too. do you think that's a good idea?
You've nailed it. Its odd your post is still hanging around at Score: 1 even though its the only one that correctly explains the glaring flaw in the benchmark.
ext2 does no journalling, so all 10 megs of your writes were buffered into ram. The benchmark then returned done, even thought little or no data had reached the disk yet.
The journalling filesystems do the same thing, with one major exception. The metadata changes MUST be written to the journal before the benchmark returns done. The data can stay in memory, since XFS and reiserfs only do metadata journalling.
You're comparing the time to write 10megs to RAM, read it back from ram and then remove it from ram against having to write 20000 files worth of metadata (10000 for the creates, another 10000 for the deletes) to disk, plus all the RAM writes/reads..
go to opensecrets.org.
MS gave just as much money (4 million) to gore as they did bush.
Why take chances when its cheap (for MS) to pay em both off?
Maybe its my database background speaking, but non-normalized (for the uninitiated read: redundantly stored, more or less) data storage is bad news. If you have redundant data, the different copies WILL eventually become out of sync.
UNIX file (1) uses the contents to determine type. Any other form of meta-data is redundantly storing the type and the contents.
Its a pain when I download a wav file on my mac and the OS thinks Netscape should open it. The solution is to not store type metadata redundantly where it can become out of sync with the real type (determined by the actual comments).
Its by one of the RHAD labs kids. Its basically just a GUI around bohem's garbage collector in leak-detector mode.
Its not purify (it really aims for leak detection, not all the other errors purify finds), but the efence + memprof combination gets you about 85% of purify's functionality.
It seems to handle threaded apps reasonably well, and C++ doesn't faze it. The only down side is that its hard to get running on non-x86 platforms.
http://ffmpeg.sourceforge.net/
Their website sucks, but the software can transcode between MP3, AC3, WAV, and RealAudio. It can also transcode video, supporting OpenDiVX, MSMPEG4 (real DIVX), MPEG2, RealVideo and MJPEG. All without stupid windows codec hacks.
Its mainly designed to be a realtime encoding and streaming system, but it includes a utility that can be used like sox.
I'm not sure why this hasn't really made a splash in the free software world yet, since the codecs included in ffmpeg mean people can play DiVX videos without any windows files, and on other CPU architectures.
Go read section 14.27 in the same RFC I refered to above. It is about "If-Range" which is the range based version of If-Match.
So no, its not different. And yes, its an obvious extension. So obvious that the HTTP/1.1 people included it.
You couldn't be more wrong.
January 1997 -- rfc2068 HTTP/1.1
See section 14.20, 14.25, 14.26, and 14.43.
It describes the "ETag: " header, which is usually a md5 hash of the resource.
The client can then validate the resources in its cache by sending a request with a "If-None-Match: " header with the ETag associated with the copy in its cache.
The server will either respond "Not modified" in which case the client simply uses the version in its cache, or the server will resend the resource if the ETags don't match.
Since this patent was filed for in 1999, this is pretty clear prior art, in the most commonly used protocol on the largest network in the world. If the patent office can't locate prior art in incredibly obvious (obvious to anyone skilled in the art, that is) cases like this one, what hope do we have for them intelligently handling more subtle cases?
actually, on my powerbook, I map the "spare" enter key (just to the right of the space bar) to the right mouse button. When using the mouse, I place my thumb on the mouse button, my pointer finger on the trackpad, and my ring finger on the enter key. I find this arrangement even better than a standard two button trackpad, since I don't have to move any fingers to hit either button.
The only bummer is that I had to map F12 as the middle click, since the extra ALT button next to the spare enter key generates the same keycode (at the OF level) as the main ALT key. Fortunately pasting is not so common that its a huge problem, but it makes using xfig a PITA.
Anyone know of a way to hack OF to make it generate different keycodes for the left and right ALT buttons? I know the ADB emulation happens there, so I'd bet its possible.
Fading mouse trails are _NOT_ motion blur. I blame 3dfx and their "T-Buffer" bullshit for this misconception. Simply mixing each frame with faded versions of the last few frames is just plain ugly. It certainly isn't motion blur.
_ motion.htm
see: http://freespace.virgin.net/hugo.elias/graphics/x
for a better description.
check out the section on file descriptors:
http://www.kegel.com/c10k.html
http://www.cs.rpi.edu/~valoisj/papers/
VSA100's motion blur isn't really motion blur. its just an accumulation buffer.
/ x_motion.htm
They just blend down the last n frames. if n = 4 you see 4 copies of the object, overlaid, each more transparent than the last. Actual motion blur is temporal antialiasing. The best way to do it is to draw each pixel n times, jittered a random amount backwards in time (a random amout of time, not a random amount of frames!).
see:
http://freespace.virgin.net/hugo.elias/graphics
near the bottom he goes into the difference between motion blur and bad fake motion blur (which is what the 3dfx card does).
(its a great page all around on graphics programming. highly recommended).
They only hand out the awards for best support on BEA newsgroups. Posting this weblogic fanboy bullshit on slashdot isn't gonna win you the big prizes.
By the way, you missed a great party on friday.
machine a: /home | nc machine.b.ip 12345
tar cvf -
machine b:
nc -l -p 12345 | tar xvf -
you can add a 'z' to the front of xvf and cvf if you want it to compress as it streams. may or may not be a good idea depending on how fast your CPUs are relative to your network.
So who here has a BOLT drive that will only store/restore a few megs with the ide-tape driver from 2.2.14?
Now that Aiwa computer products is no more, what are the chances of us ever getting the info to support these things?
If you're bored, call:
1-800-321-AIWA (1-800-321-2492) option 5
and ask for the info needed to support this discontinued hardware on linux and BSD.
Add the following to your .Xdefaults
Netscape*toolBar.search.isEnabled: false
Netscape*toolBar.destinations.isEnabled: false
Netscape*toolBar.myshopping.isEnabled: false
If you'd like a "Find" button to search for text in the current page/frame add:
Netscape*toolBar.numUserCommands: 1
Netscape*toolBar.userCommand1.commandName: findInObject
Netscape*toolBar.userCommand1.labelString: Find
Netscape*toolBar.userCommand1.commandIcon: Find
Enjoy.
Type "info gcc" and look at all those options. most of those change the way gcc generates machine code given a certain chunk of source.
And that is just one compiler, there are many. There is a reason that C decompilers are rare and don't work well.