Check this work unit out....

http://setiweb.ssl.berkeley.edu/workunit.php?wuid=14743107

There are 2 hosts with download errors. If you check the wu detail it says they are still on client 4.13, hence they got the errors. Isn't it about time admin set the min client level to 4.19 and above?

Live long and crunch!

---

Paul
(S@H1 8888)
And proud of it!

<blockquote>Check this work unit out....

http://setiweb.ssl.berkeley.edu/workunit.php?wuid=14743107

There are 2 hosts with download errors. If you check the wu detail it says they are still on client 4.13, hence they got the errors. Isn't it about time admin set the min client level to 4.19 and above?

Live long and crunch!
</blockquote>

It doesn't work quite as you suggest.
The science app can require a minimum version of BOINC.
The data can specify a required version of the science app.
But there is no direct way for the data to require a minimum version of BOINC.

When Berkeley release a new version of the science app it can demand a new BOINC. But, I'm sure Berkeley will do all they can NOT to release an App that requires a new BOINC too - It may happen at version 5 - even then many will hope not - those with farms or systems they don't have regular acces to will hate it if their systems suddenly stop working until they can re-install.

Besides, a download client error doesn't really cause anyone a problem as the failed result will be re-allocated and downloaded to another host soon - in this case a little over 24 hours later.

---

Oh, I've seen them set the min client level before easily enough. It has been a while though.

---

Paul
(S@H1 8888)
And proud of it!

<blockquote>Oh, I've seen them set the min client level before easily enough. It has been a while though.
</blockquote>

Yes, no argument there, the science app can require a minimum version of BOINC (more specifically a given Major version number e.g. a 4.x app needs a 4.x BOINC IIRC)

The point is no one wants to release such an app - not for an existing project anyway. It would cause mayhem if everyones systems stopped working because a new app aoto-downloaded and nothing would run because the framework (BOINC) is not current. BOINC itsself does not auto-update (thank heavens).
When you're attaching to a new project, then it's not so much of a pain to require a new BOINC install.

---

Na. I'm with GG on this one. All projects should now require BOINC 4.19 minimum.

---

Kolch - Crunching for the BOINC@Australia team since July 2004.
Search for your own intelligence...

<blockquote>Na. I'm with GG on this one. All projects should now require BOINC 4.19 minimum.
</blockquote>

Maybe I'm missing something, but I'm not sure I follow why it matters to participant A what version of BOINC participant B is running - B may be experiencing problems, missing functionality and losing even work, but that is their loss. The science is not affected by the BOINC version.

AFAIK, having a bunch of users on older versions isnt holding back project development. If a new project comes along that does need a later version and A.N.Other wants to run the new project they'll have to upgrade.

---

I just posted this under questions/problems but I keep getting this message (although for different WU's):

5/12/2005 1:52:23 PM|SETI@home|Starting result 19dc04aa.17564.26192.615894.28_0 using setiathome version 4.09

I just started Boinc with 4.19 in March and upgraded to 4.37 last week during the outage. It's not working out as well as I had hoped.

---

<blockquote>I just posted this under questions/problems but I keep getting this message (although for different WU's):

5/12/2005 1:52:23 PM|SETI@home|Starting result 19dc04aa.17564.26192.615894.28_0 using setiathome version 4.09

I just started Boinc with 4.19 in March and upgraded to 4.37 last week during the outage. It's not working out as well as I had hoped.</blockquote>

Kathy,
I think you are confusing the BOINC and SETI application versions.
The current SETI application is 4.09.
The current stable release of BOINC is 4.25, and the latest beta (test) version is 4.38
You are running the correct version of SETI@Home, and verion 4.37 of BOINC is very recent, but 4.38 exists, and 4.39 can be expected soon.
If 4.37 is working for you I would stay with it as 4.38 has its own problems.

---

<blockquote>I just posted this under questions/problems but I keep getting this message (although for different WU's):

5/12/2005 1:52:23 PM|SETI@home|Starting result 19dc04aa.17564.26192.615894.28_0 using setiathome version 4.09

I just started Boinc with 4.19 in March and upgraded to 4.37 last week during the outage. It's not working out as well as I had hoped.</blockquote>
Kathy,

This is a normal process/status message that you will see as the BOINC Manager changes between work units. You can look up all the messages I know about in the BOINC Owner's Manual. Look and click on my signature image ...

---

Some of us are just lazy, or paranoid, or leary of problems with upgrades to do it. I only just got off my tush and upgraded from 4.19 to 4.25. (Personally, I was worried about losing a CPDN unit, but, all went well.)

---

<blockquote><blockquote>Na. I'm with GG on this one. All projects should now require BOINC 4.19 minimum.
</blockquote>

Maybe I'm missing something, but I'm not sure I follow why it matters to participant A what version of BOINC participant B is running - B may be experiencing problems, missing functionality and losing even work, but that is their loss. The science is not affected by the BOINC version.

AFAIK, having a bunch of users on older versions isnt holding back project development. If a new project comes along that does need a later version and A.N.Other wants to run the new project they'll have to upgrade.

</blockquote>
Because now you HAVE to wait for the 3rd person to return the unit and HOPE that they will not have a problem or you WILL be waitng until they find someone that WILL return the unit. You KNOW that the person with the problem computer is STILL getting units and NOT going to return ANY of them sucessfully!

---

<blockquote><blockquote>Na. I'm with GG on this one. All projects should now require BOINC 4.19 minimum.
</blockquote>

Maybe I'm missing something, but I'm not sure I follow why it matters to participant A what version of BOINC participant B is running - B may be experiencing problems, missing functionality and losing even work, but that is their loss. The science is not affected by the BOINC version.

</blockquote>



I know of at least one WU that where 6 4.13 hosts "lost" a WU so it was thrown out. This not only hurts the science but is a waste of time for the 2 hosts that finished the WU...and will never get credit for it.

Don

---

<blockquote><blockquote><blockquote>Na. I'm with GG on this one. All projects should now require BOINC 4.19 minimum.
</blockquote>

Maybe I'm missing something, but I'm not sure I follow why it matters to participant A what version of BOINC participant B is running - B may be experiencing problems, missing functionality and losing even work, but that is their loss. The science is not affected by the BOINC version.

AFAIK, having a bunch of users on older versions isnt holding back project development. If a new project comes along that does need a later version and A.N.Other wants to run the new project they'll have to upgrade.

</blockquote>
Because now you HAVE to wait for the 3rd person to return the unit and HOPE that they will not have a problem or you WILL be waitng until they find someone that WILL return the unit. You KNOW that the person with the problem computer is STILL getting units and NOT going to return ANY of them sucessfully!
</blockquote>

I see where you're comming from.

Doesn't bother me in the least, I know the credit will come. I just reached 100,000 credits, and I have maybe 300-500 pending. No big deal.
When I reach 200,000 I'll still have 300-500 pending, but not the same ones.

It's a bad reason for wanting people to update - where does it lead?
No-one should run beta-test versions - as some WUs always lost get in the switch-over or through failure?
No-one with a CPU that takes more than 4 hours per WU sould be allowed - too long to wait?
No cache bigger than .1 day - the WU you're waiting for might be at the end of someones 7 day cache?

---

Thought I'd chime in here. When a "Download Error" occurs, the client knows this, and the next time it reports to the Data server (berkeley) it will be reported and a another copy of the WU will be dispatched in short order. I've seen that more times than most the Copy will reissued the same or next day of the first release. Down load errors don't really delay the issuance of credit by more than a day in 90% of cases. My study of 187 WUs shows:

Number of results sent out = 907
Number of D/L errors = 106
Number of "no reply" errors = 58
Total errors = 182
Average number of days a result takes from the time it's issued until the last date of return (to seti)= 9.07 (including first and last day)

Again, D/L errors don't really delay the issuance of credit by more than a day, the real killers are the "over no reply" errors, since it takes the full two weeks before another copy is issued.

tony

[edit] I've sorted the data by D/L error and then by No reply error.

the Average number of days for results with D/L errors is 10.36 days, and the average for those with "no reply" is 12.41 days.

Download errors increase the number of days from 9.07 to 10.36 or by 1.29 days.

You might wonder why the No Reply days is less than the 14 days the error must exist, well, that's because 4 identical results are initially sent, but only 3 are required to validate, and that's what's happening. the 4th WU isn't needed.

Also, you might think the Avg # of days from issuance to return means thats how long you have to wait for credit, again, only 3 are required for the quorum, so you get your credits a bit faster than the number of days.

---

Thanks so much, Mike! That's exactly it.

Kathy

---

I think that they can require a minimum specific release number... from the Einstein page: "[Use version 4.19 or greater!]"

I don't know what happens if you try to connect with a lower version, but it at least seems as though they can require a certain version...

---

<blockquote>Maybe I'm missing something, but I'm not sure I follow why it matters to participant A what version of BOINC participant B is running - B may be experiencing problems, missing functionality and losing even work, but that is their loss. The science is not affected by the BOINC version.
</blockquote>
4.13 had some download issues.

---

<blockquote>I think that they can require a minimum specific release number... from the Einstein page: "[Use version 4.19 or greater!]"

I don't know what happens if you try to connect with a lower version, but it at least seems as though they can require a certain version...</blockquote>
If you try to connect with a release lower than the one required, you get a message telling you you need to upgrade the Boinc Client. No work is sent out to you until you do.

---

Anyhow its also an issue of efficiency. The electrical costs of running PC's flat out crunching WU's. So errors should be avoided. Yes I also think that if a PC requires more than 24hours to crunch a WU that the user should seriously consider turning it off, or finding a less computationally demanding distributed computing project.

---

Kolch - Crunching for the BOINC@Australia team since July 2004.
Search for your own intelligence...

Thanks too, Paul! I use it a lot. :)

---