Consider forcing old clients to DC?

[leonardyan96]

When PRP proof finally gets formally available, traditional first primality tests produced by old clients would become somewhat a waste of horsepower. Would it be sensible to force clients < v30 to DC? Here's some of my thoughts.
Starting from v30 getting released, we can wait during a tolerance period of several months even a year. After that period, when old clients before v30 ask for LL/PRP first test assignments, PrimeNet will return LL/PRP-DC assignments respectively. If one day there's no more exponents for DC, we can still give PM1/ECM to remaining old clients, if there will be any.
Also, when these clients contact the server, maybe PrimeNet could explain why and remind these contributors of upgrading software by returning messages?

[retina]

Originally Posted by leonardyan96

...force clients ...

Welcome to the world of Internet giants where forcing the suckers customers to do what they want is the norm.

No! Let's not force anyone into doing anything. Just have the work dry up and let the user decide if they wish to change work type or quit.

[ric]

Originally Posted by retina

let the user decide

A rather subtle difference, I'd say

[xilman]

Originally Posted by retina

Welcome to the world of Internet giants where forcing the suckers customers to do what they want is the norm.

No! Let's not force anyone into doing anything. Just have the work dry up and let the user decide if they wish to change work type or quit.

How nice to see someone employing irony.

[storm5510]

Originally Posted by leonardyan96

When PRP proof finally gets formally available, traditional first primality tests produced by old clients would become somewhat a waste of horsepower. Would it be sensible to force clients < v30 to DC? Here's some of my thoughts.

Starting from v30 getting released, we can wait during a tolerance period of several months even a year. After that period, when old clients before v30 ask for LL/PRP first test assignments, PrimeNet will return LL/PRP-DC assignments respectively. If one day there's no more exponents for DC, we can still give PM1/ECM to remaining old clients, if there will be any.

Also, when these clients contact the server, maybe PrimeNet could explain why and remind these contributors of upgrading software by returning messages?

The simple solution would be to present a short line of text in the Comm thread of older clients saying LL, LL-DC, and PRP-DC no longer exist. Then, recommend the update to v30. All remaining LL-DC work could be ran as first-time PRP's to generate the certification data. None of this would need to happen until v30 goes "public." That is, being shown as available on Priment's download page. Many users are not forum members. They have no idea any of this is happening.

I know George, a.k.a. "Prime95," is working on solutions for all of this. v30 is a major update. It will take time to work out all of the kinks. The best way to keep up with the revisions is here as they come along.

[Aramis Wyler]

I don't think there is any harm in having the older clients to LLDC work. Running the DCs as PRPs doesn't save us much, and that's a lot of cycles that would be shut off.

I believe there is a fallback work type for most of those versions if there is none of the regular work, so letting the LLFT dry up would automatically trigger LLDCs for anyone who had that or "let primenet decide" as their backup work.

And that would be abiding by the user's wishes, as well.


So to sum up, I think if we simply stop releasing LLFT numbers, most of the <v30 clients will automatically cut over to LLDC which is not a bad thing.

[S485122]

Originally Posted by Aramis Wyler

I don't think there is any harm in having the older clients to LLDC work. Running the DCs as PRPs doesn't save us much, and that's a lot of cycles that would be shut off.

I believe there is a fallback work type for most of those versions if there is none of the regular work, so letting the LLFT dry up would automatically trigger LLDCs for anyone who had that or "let primenet decide" as their backup work.

And that would be abiding by the user's wishes, as well.


So to sum up, I think if we simply stop releasing LLFT numbers, most of the <v30 clients will automatically cut over to LLDC which is not a bad thing.

Indeed !
Jacob

[leonardyan96]

Originally Posted by Aramis Wyler

I don't think there is any harm in having the older clients to LLDC work. Running the DCs as PRPs doesn't save us much, and that's a lot of cycles that would be shut off.

I believe there is a fallback work type for most of those versions if there is none of the regular work, so letting the LLFT dry up would automatically trigger LLDCs for anyone who had that or "let primenet decide" as their backup work.

And that would be abiding by the user's wishes, as well.


So to sum up, I think if we simply stop releasing LLFT numbers, most of the <v30 clients will automatically cut over to LLDC which is not a bad thing.

This consequence is same as what I personally desire.

[leonardyan96]

Originally Posted by storm5510

The simple solution would be to present a short line of text in the Comm thread of older clients saying LL, LL-DC, and PRP-DC no longer exist. Then, recommend the update to v30. All remaining LL-DC work could be ran as first-time PRP's to generate the certification data. None of this would need to happen until v30 goes "public." That is, being shown as available on Priment's download page. Many users are not forum members. They have no idea any of this is happening.

I know George, a.k.a. "Prime95," is working on solutions for all of this. v30 is a major update. It will take time to work out all of the kinks. The best way to keep up with the revisions is here as they come along.

What I worry about is that, lots of average participants may never read p95 screen outputs, especially for those clients running on remote servers. The solution described in #6 looks nice.