Printable View
According to the following article, Klaus is planing on creating CD and DVD versions of Knoppix for future release.
http://news.zdnet.co.uk/software/lin...9181773,00.htm
Acording to the article, "The Knoppix Linux Live CD will be split into two editions - a 'light' version for older computers, and a 'maximum' version that has all the Debian packages"
Will that light Version be the same as whats out now? Or even less packages.
If the latter I hope a normal CD is still made.
I'm not sure, but it sounds like that the CD will remain pretty much as it is. The thing that seems to be new is the addition of a DVD which I suppose will be published whenever a new CD is published.Quote:
Originally Posted by firebyrd10
Imagine the poor schlub that tries to pull the DVD down via dial-up... :shock: You know someone will try to do just that. Still, even on broadband that download should take quite a while, relatively speaking.
'Goon
Twelve euros to iso-top.info is so much more practical than 72 hours on broadband to New Zealand. Both work, though.
I suppose the next job is to eliminate the 2GB 'line' that the KNOPPIX file seems to suffer with, and figure out what to put on the 50GB BluRay DVDs that we get next year. And the 25GB biodegradeable corn-starch ones, assuming the news article wasn't a joke.
I put a 'qemu' with a Knoppix. The result of that is something that autoplays under Windows like http://unit.aist.go.jp/it/knoppix/qemu/index-en.html . It works, too.
You can also do this with games-knoppix from http://games-knoppix.unix-ag.uni-kl.de/ . St. Nicholas has the ability and the permission to distribute an unlimited number of these on Friday night (24/25 Dec), and it would make a great present for lots of kids if he did. I wonder if he will.
It looks like probono has already done that with the cmg format packages he is using. Knoppix can have a 2 gig file, then plenty of more applications to fill the rest of the disc (even double layer discs until we get those blue ray drives) in cmg files.Quote:
Originally Posted by chris-uk
How about SFS-KNOPPIX to use 'light' and 'maximum' knoppix?
http://unit.aist.go.jp/itri/knoppix/sfs/index-en.html
SFS-KNOPPIX is independent of cloop-size, because it share cloop files via Internet with SFS(Self-certyfing File System) and downloads requested blocks only.
CD of SFS-KNOPPIX includes boot loader only. We can select 'light' and 'maximum' knoppix at boot time.
SFS-KNOPPIX is discussed in the following forum. Please join.
http://www.knoppix.net/forum/viewtopic.php?t=14904
What is needed for SFS KNOPPIX is a server in this country to reduce the inherent delay from far away Japan.
SFS KNOPPIX operates well and does permit the use of apps greater than the CD limits up to the servers limits of stored programs.
Maybe someone in the USA can provide the server?
I guess SFS Knoppix is OK for what it is, but I certainly hope I don't see it find it's way into the regular release CD or DVD. I hope the DVD is full of applications (well beyond the 2 gig limit by using cmg files as well as the clopp). I want to be able to use it on an airplane, or when I'm at my parents home with no Internet access.Quote:
Originally Posted by lilsirecho
SFS looks like it could be very handy if one has a local SFS server. By local I mean on one's own loacl network. It's less clear to me that a server in my country would be a big improvement (although it would likely help, but maybe not - I used get my new Knoppix ISo files before I used bittorrent from a server off my continent, because all of my national servers were much slower than some foregin servers.)
Harry;
The SFS KNOPPIX full CD download is obtained in 3 minutes in Japan (Japan is 1200 miles long).
The download of the full CD from Japan to USA at 88kB/sec is a 30 minute task.
The improvement is great for the full CD.
The download for any part of the full complement is proportional to its size and the download speed of the internet connection. Thus, access to many programs is provided by the initial 20MB download, including those supposedly DVD only.
I question the use of cmg in tandem or conjunction with Knoppix since knoppix operates on a different basis(uncompressing on the fly).
I am not certain that the program k3b permits additions to knoppix installed CD's since the mechanism utilized for install in such a case is isolinux as initiated by the program selected in k3b.
Whether additional material can be added which is compatible with the knoppix format is certainly not clear.
Probably material to fill the DVD available space is possible with another program and a different access mechanism but can't see the advantage off-hand. It means restarting for access.
Perhaps I do not understand cmg sufficiently to be more specific but do appreciate the requirements for burning Knoppix to CD or DVD and the consequent restrictions.
Well, from that it does seem that SFS is viable in Japan. It is less clear that just having a server in the US would speed things up in the US. It could well be that Japan has a better infrastructure. The US is certainly playing catch-up in many areas, I don't know if just having a US server would speed up SFS for all US users. Maybe it would, but I do have doubts. We also have a great ability to overload what servers we do have (just look at the US Knoppix mirrors). So if SFS was speed viable in this country we might then move on to the problem of it being too sucessful, and everything grinding to a halt. An SFS server on the local netwirk still strikes me as the best way to go, but I'll certainly acknowledge this is an outsiders viewpoint, while I've been following the discussions on SFS I have not tried to use it.Quote:
Originally Posted by lilsirecho
Harry;
Pleased to hear from you on the subject of SFS.
Indeed, the time delay on internet communications from overseas can be likened to the time delay(on a greater scale) of satellite telephone links, or TV for that matter.
The delay from coast to coast can be compared easily with the delay to Japan by using the ping parameter.
Within the USA, I would expect the 3 minute download time on broadband (768kb/s) to be very comparable.
Most assuredly, the local network server would be ideal. However, when one is only downloading one application, the time becomes quite short, often ten to twenty seconds.
The CD to enable this SFS arrangement is only 20MB in size.
The downloads are not .iso files needing a CD burn but are cloop files decoded "on-the-fly" so there is not any CD or DVD writer needed after the initial burn of 20MB. All files are kernel related so a selection is made at download time.
I suggest you give SFS a try and while you do that, ping their IP to see the delay across the pond and compare it to across the USA pings.
The system seems to have extralarge capability and might be best applied with servers dedicated to specific kernels and maybe....no kernels at all just cloop's.
.................................................. ..
Doc, I have a serious memory problem!!
When did you first notice this problem?
What problem?
Harry and lilsirecho,
Please tell me the result of "ping ring.aist.go.jp". The time will take more than 100 msec.
Network latency is important factor to realize SFS-KNOPPIX, because every access to cloop have to take the time of latency.
The time of "ping ring.aist.go.jp" from my home ADSL is 20msec. The boot time of SFS-KNOPPIX is about 5 minutes.
1 minute of mentioned 5 minutes is spent to prepare SFS at first and 2 minutes is spent for KDE.
The bandwidth of network is 12Mbps (in catalog) but 5Mbps (in real).
Instead of MIRROR SFS server, we can use "sfsrocd(sfs read only client daemon)".
"sfsrocd" relay download blocks of SFS.
http://www.pdos.lcs.mit.edu/papers/sfsro.html
http://www.usenix.org/events/osdi200...pers/fu/fu.pdf
"sfsrocd" may solve the network latency problem, if there are many SFS-KNOPPIX users. :-)
[root@n6re ~]# ping ring.aist.go.jp
PING ring.aist.go.jp (150.29.9.6) 56(84) bytes of data.
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=1 ttl=237 time=280 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=2 ttl=237 time=281 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=3 ttl=237 time=300 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=4 ttl=237 time=283 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=5 ttl=237 time=281 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=6 ttl=237 time=283 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=7 ttl=237 time=281 ms
64 bytes from ring.aist.go.jp (150.29.9.6): icmp_seq=8 ttl=237 time=280 ms
--- ring.aist.go.jp ping statistics ---
9 packets transmitted, 8 received, 11% packet loss, time 8009ms
rtt min/avg/max/mdev = 280.848/284.166/300.535/6.297 ms
[root@n6re ~]#
sure.Quote:
Originally Posted by suzaki
Of course I realize network latency is an issue (although there are ways to work around this. If one doesn't wait for every packet and every acknowledgement before requesting and sending the next packet, the problem goes away (a.k.a. sliding window). It's also important to not place too much importance on a ping; while it's a starting point, not all packets move at the same speed and a ping may move through a network a lot faster than some other packets.Code:Pinging ring.aist.go.jp [150.29.9.6] with 32 bytes of data:
Reply from 150.29.9.6: bytes=32 time=249ms TTL=237
Reply from 150.29.9.6: bytes=32 time=250ms TTL=237
Reply from 150.29.9.6: bytes=32 time=250ms TTL=237
Reply from 150.29.9.6: bytes=32 time=249ms TTL=237
Ping statistics for 150.29.9.6:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 249ms, Maximum = 250ms, Average = 249ms
But the real issue here is not just how long it takes me to ping the site in Japan, but how long it would take for someone in the United States to reach a site in the United States. What ping speeds are you seeing for your site from Japan? I know that in the United States I often see pings nearly 200 ms for servers in this country, and not infrequently a ping of a server in the US may take even longer than the above ping times to Japan. So as I say, it may be a matter of our infrastructure just not being up to the standards that yours is. Servers in Japan might serve SFS in Japan quite well, but putting a server in the U.S. may not solve all speed problems for U.S. users.
Harry,
As you say, SFS-KNOPPIX is not goot at long network connection.
One way to solve this problem is to replace SFS with SFSRO.
http://www.pdos.lcs.mit.edu/papers/sfsro.html
http://www.usenix.org/events/osdi200...pers/fu/fu.pdf
"sfsrocd(sfs read only client daemon)" relays donwnloaded blocks to another client.
It works as P2P.
By the way, here is forum for "Knoppix to split into CD and DVD versions". Please move to forum for "SFS-KNOPPIX".
http://www.knoppix.net/forum/viewtopic.php?t=14904
Hi again,Quote:
Originally Posted by suzaki
Have you given any thought to a "sliding window" approach rather than waiting for each packet to complete before starting the request for the next packet? I think this would have the largest and a very significant impact on the latency problem.
I'm not at all clear on what you are requesting me to move. The link that you gave is for one of your original SFS-KNOPPIX posts. It was posted in the news forum. This thread is also in the news forum. I could move that thread and/or this thread to any other existing forum that you like (such as the lounge or tips and tricks), but there is no SFS-KNOPPIX forum. I cannot create one. If you expect a forum just for SFS-KNOPPIX you would have to talk to eadz, but I don't expect that there can or should be a new forum created for each new variant of Knoppix. I cannot combine the two threads into one thread. I'm completely confused by the part that reads: By the way, here is forum for "Knoppix to split into CD and DVD versions". So if there is something else that you were hoping I would do for you, please try to restate the request.
Please tell me how to setup sliding window.Quote:
Originally Posted by Harry Kuhman
I just think our discussion is much suitable topic of "SFS-KNOPPIX" in News forum.Quote:
Originally Posted by Harry Kuhman
Actually, since TCP/IP is intended to let packets arrive out of order and put them together in order, I would have thought that this happens by default. But you seem to have a system that is very very dependent on the latency of every packet, which implies that you are waiting for the acknowledgement on each packet before sending the next. I'll see what I can learn and post back, but it may be something you have to look at closer at your end and in your code.Quote:
Originally Posted by suzaki
Well, the original announcement certainly fit in the News forum. This thread is in News also. Personally I would think you might want to move it to Lounge, but I'm not trying to tell you to do that, this may be as good as a place for SFS-K discussion. What I don't understand is what you are asking me to do, what you were intending when you posted addressed to me saying "By the way, here is forum for "Knoppix to split into CD and DVD versions". Please move to forum for "SFS-KNOPPIX". ". If not in the news forum, where we currently are, where would you want me to move the thread that you pointed me to?Quote:
Originally Posted by suzaki