There is an issue here to do with the servers initiating connections with the host machines which seems to has been forgotten. Firewalls.
I think the vast majority of people try to limit incoming connections to the bare minimum. Others are more obsessive and ban all incoming connections not initiated by their computer. And this is without talking about some Windows firewalls which are just plain paranoid, and also make it extremely difficult to change their settings.
I guess what I am trying to say is that it would be impossible to contact some host machines, and the owners of others would start inundating the boards with compliants about the servers being 'hacked' and trying to 'infect' their machines. The users at Einstein seem to have a bit more know-how, so it wouldnt be as bad as at SETI/BOINC. Remember what happened when the BOINC GUI RPC calls (I think thats what they are) were changed when the 4.2x series came out?
If this were implemented, I think it would have to be in the client.
Personally, I don't really like this idea. I am also someone who can't afford to run his computer 24/7 and upgrade every 6 months. Even if I was given credit for a result I was crunching which was declared 'redundant', I would feel that my contribution was being ignored. It would probably also penalise people who contribute to more than one project.
I hope I explained what I am trying to say, I'm not too good on tech speak, so I've propably got a lot of the facts wrong as well
Edit, Spelling etc
Need Help? Try the excellent Unofficial BOINC Wiki!
We are the BOINC. Prepare to be assimilated.
'anthrax beats WinXP' - The Register
This thread is pretty old, I had forgotten about it. The small amount of efficiency gained is surely outweighed by having another option for users. The client/setup should remain as simple as possible. It is already the case that people are confused by it. Adding an extra option in this case is unjustified.
The server does not initiate contact. This would all take place within the context of the client-initiated session. However, it is true that it would only be useful for PC's which contacted the server every work unit. Not everyone has an always-on connection, so obviously the return would be very small.
Quote:
Even if I was given credit for a result I was crunching which was declared 'redundant', I would feel that my contribution was being ignored.
I absolutely agree. No-one's work is redundant, because if not for a slower host, faster hosts would then be allocated to do the fourth result. Allocating the hosts with greater connection intervals to the fourth slot improves efficiency. They are making the system more efficient, and I don't mean to overlook that. All users' work is useful and makes for a successful distributed computing application. Users should rightly feel proud when they receive credit, they have earned it and it is appreciated.
There is an issue here to do
)
There is an issue here to do with the servers initiating connections with the host machines which seems to has been forgotten. Firewalls.
I think the vast majority of people try to limit incoming connections to the bare minimum. Others are more obsessive and ban all incoming connections not initiated by their computer. And this is without talking about some Windows firewalls which are just plain paranoid, and also make it extremely difficult to change their settings.
I guess what I am trying to say is that it would be impossible to contact some host machines, and the owners of others would start inundating the boards with compliants about the servers being 'hacked' and trying to 'infect' their machines. The users at Einstein seem to have a bit more know-how, so it wouldnt be as bad as at SETI/BOINC. Remember what happened when the BOINC GUI RPC calls (I think thats what they are) were changed when the 4.2x series came out?
If this were implemented, I think it would have to be in the client.
Personally, I don't really like this idea. I am also someone who can't afford to run his computer 24/7 and upgrade every 6 months. Even if I was given credit for a result I was crunching which was declared 'redundant', I would feel that my contribution was being ignored. It would probably also penalise people who contribute to more than one project.
I hope I explained what I am trying to say, I'm not too good on tech speak, so I've propably got a lot of the facts wrong as well
Edit, Spelling etc
Need Help? Try the excellent Unofficial BOINC Wiki!
We are the BOINC. Prepare to be assimilated.
'anthrax beats WinXP' - The Register
This thread is pretty old, I
)
This thread is pretty old, I had forgotten about it. The small amount of efficiency gained is surely outweighed by having another option for users. The client/setup should remain as simple as possible. It is already the case that people are confused by it. Adding an extra option in this case is unjustified.
The server does not initiate contact. This would all take place within the context of the client-initiated session. However, it is true that it would only be useful for PC's which contacted the server every work unit. Not everyone has an always-on connection, so obviously the return would be very small.
I absolutely agree. No-one's work is redundant, because if not for a slower host, faster hosts would then be allocated to do the fourth result. Allocating the hosts with greater connection intervals to the fourth slot improves efficiency. They are making the system more efficient, and I don't mean to overlook that. All users' work is useful and makes for a successful distributed computing application. Users should rightly feel proud when they receive credit, they have earned it and it is appreciated.