If routing tables are indexed search times only grow at log(n) where n is the search time. Memory is cheap would a gig of memory in a device be a bad thing? A standard desktop machine currently has 128M. This will give the ability to accomodate a couple of order of magnitude more routes in the routing table easily. If we go down the road of geographically based routing governments will finally have a throat to squeeze and will be as happy as a dog with two dicks and the real power of the internet will have been destroyed. Enable security but don't weaken the architecture.
If he negotiated copyright ownership as part of the employment contract, then he owns it simple. If this is on paper it is far firmer than an oral agreement.
Point 2 depends on 1 you if he is the copyright owner and he released the code all is well. If he isn't the owner then he has no right to release the code under the GPL.
Integrate your account management into your NT domain or existing domain. There is a new paper somewhere on http://www.samba.org explaining how to do this for NT domains. Managing multiple accounts is a pain in the arse. If the system is virtually invisable to your system admins they will become very positive about linux.
Ram and Scramjets make sense for atmospheric lift
on
X-33 Shuttle Problems
·
· Score: 1
Spacecraft could get through the $1000 pound barrier if they simply used ramjets and cut back on the LOX. Has anyone done any testing recently on the maximum speed of a Hyd ramjet. Mach 10 should be possible which is a great head start and there has been speculation about mach 25. If the Germans could get ramjets flying fueled with coal dust in the second world war we should be able to do a little better today.
I like linux and earn a living doing unix administration. I also do a little microsoft admin and recently I went on an Win2000 course. Nothing in these courses gives you any real idea of server performance but there is a lot in 2000 that should be looked at by the linux community. Active directory, secure dynamic DNS updates and the degree of integration of security. Unix in general has had a long time to get these features right and the results are NIS+ (vomitus muchus) DCE etc.
What is wanted is a great kerberos or better authentification implimentation that can use a variety of encryptional algorithms (heimdal(?)), a secure directory server architecture openldap2.0(?) and office applications that run from the server so we don't have any of this client configuration bullshit and all personal config is pulled from a directory service. The inability to break away from client has made microsoft pursue extremely complex solutions when very simple solutions will do. Client config and applications should be kept to a minimum it should be just a grunt bucket with lots of ram.
All a distrubuted OS does is make a cluster more managable, just an OS make the disparate parts of your PC manageable. But it all only works if your problem domain is suitable to being broken up in a parellel manner and communications between the nodes is adequate for your problem. If you are breeding a flock/herd/whatever of geneitic algorithms mosix is probably the best choice because the algorithms only need to interact to mate and cull. PVM is best where there is a greater need for interprocess communication, MPI is less friendly in so far as it doesn't give you the virtual machine but it can apply greater rigor and is a standard parellel programing enviroment. This isn't to say that a combination of mosix and PVM can't be used, they can. Some other things to look at are shrimp. Currently one size doesn't fit all because different problems create different demands on the infrastructure. Does the solution need much IPC, how big are these messages? Ethernet is usually involved in cheap clusters it should be seen as a slow bus that can carry large messages.
Design it properly from the start. Use bootprom PXE to boot from the network card theres a linux howto on this. Cached local system etc update just require the PC to be rebooted. New PCs just need to be plugged into the network. Use LDAP as a directory service so mail, userinfo and password file contents are managed centrally. Use Kerberos to manage key exchange and make local caches of this info. Minimise what is kept locally.
Its may not be delusional. The recently genengineered smart mice are created simple by reactivating a gene which is usually switched off in adults. Another pointer may be found in revolutionary science Newton, Einstien and the quantum germans whose name eludes me at the moment. Weren't they all remarkably young when they made their greatest contribution? Now try to name a few old scientists who revolutionised science.........
Slashdot Mirror
If routing tables are indexed search times only grow at log(n) where n is the search time. Memory is cheap would a gig of memory in a device be a bad thing? A standard desktop machine currently has 128M. This will give the ability to accomodate a couple of order of magnitude more routes in the routing table easily. If we go down the road of geographically based routing governments will finally have a throat to squeeze and will be as happy as a dog with two dicks and the real power of the internet will have been destroyed. Enable security but don't weaken the architecture.
If he negotiated copyright ownership as part of the employment contract, then he owns it simple. If this is on paper it is far firmer than an oral agreement. Point 2 depends on 1 you if he is the copyright owner and he released the code all is well. If he isn't the owner then he has no right to release the code under the GPL.
Integrate your account management into your NT domain or existing domain. There is a new paper somewhere on http://www.samba.org explaining how to do this for NT domains. Managing multiple accounts is a pain in the arse. If the system is virtually invisable to your system admins they will become very positive about linux.
Spacecraft could get through the $1000 pound barrier if they simply used ramjets and cut back on the LOX. Has anyone done any testing recently on the maximum speed of a Hyd ramjet. Mach 10 should be possible which is a great head start and there has been speculation about mach 25. If the Germans could get ramjets flying fueled with coal dust in the second world war we should be able to do a little better today.
Doesn't the military currently limit GPS implimentations to speeds below 900km/h?
Who said that defence research was a waste of time. http://cvw.mitre.org/
Both integrate with mutt and its easy to do so.
I like linux and earn a living doing unix administration. I also do a little microsoft admin and recently I went on an Win2000 course. Nothing in these courses gives you any real idea of server performance but there is a lot in 2000 that should be looked at by the linux community. Active directory, secure dynamic DNS updates and the degree of integration of security. Unix in general has had a long time to get these features right and the results are NIS+ (vomitus muchus) DCE etc. What is wanted is a great kerberos or better authentification implimentation that can use a variety of encryptional algorithms (heimdal(?)), a secure directory server architecture openldap2.0(?) and office applications that run from the server so we don't have any of this client configuration bullshit and all personal config is pulled from a directory service. The inability to break away from client has made microsoft pursue extremely complex solutions when very simple solutions will do. Client config and applications should be kept to a minimum it should be just a grunt bucket with lots of ram.
All a distrubuted OS does is make a cluster more managable, just an OS make the disparate parts of your PC manageable. But it all only works if your problem domain is suitable to being broken up in a parellel manner and communications between the nodes is adequate for your problem. If you are breeding a flock/herd/whatever of geneitic algorithms mosix is probably the best choice because the algorithms only need to interact to mate and cull. PVM is best where there is a greater need for interprocess communication, MPI is less friendly in so far as it doesn't give you the virtual machine but it can apply greater rigor and is a standard parellel programing enviroment. This isn't to say that a combination of mosix and PVM can't be used, they can. Some other things to look at are shrimp. Currently one size doesn't fit all because different problems create different demands on the infrastructure. Does the solution need much IPC, how big are these messages? Ethernet is usually involved in cheap clusters it should be seen as a slow bus that can carry large messages.
Design it properly from the start. Use bootprom PXE to boot from the network card theres a linux howto on this. Cached local system etc update just require the PC to be rebooted. New PCs just need to be plugged into the network. Use LDAP as a directory service so mail, userinfo and password file contents are managed centrally. Use Kerberos to manage key exchange and make local caches of this info. Minimise what is kept locally.
Its may not be delusional. The recently genengineered smart mice are created simple by reactivating a gene which is usually switched off in adults. Another pointer may be found in revolutionary science Newton, Einstien and the quantum germans whose name eludes me at the moment. Weren't they all remarkably young when they made their greatest contribution? Now try to name a few old scientists who revolutionised science .........