I've just found another item on XP, from a news item about updating Adobe Flash:

And there are still so darn many Windows XP users�an estimated 250 million�and they�re not all retired people with nothing but pictures of pets and grandkids on their aging tower computers.

It�s been more than a year since Microsoft officially ended support for XP, but last month the United States Navy signed a USD$9.1 million contract with Microsoft to continue receiving extended support for XP (security patches) for any of its 100,000 computer still running the obsolete operating system as it continues to migrate its systems to a newer Windows platform.

I should include a link, as it also talks about Chinese phishing attacks, although it may not last long:
Homeless in Vancouver: Need Adobe Flash Player? It needs to be updated�again

Hi folks,

While compatibility with newer software is important, what I find interesting is that 96% ot the climate prediction crunchers are dormant, absent, inactive, lost etc. and as little as 11 183 users are crunching.

Users 285,260
Active users 11,183 (3.92%)
Hosts 599,019
Active hosts 14,735 (2.46%)

http://boincstats.com/en/stats/2/project/detail/overview
and http://boincstats.com/en/stats/2/project/detail/user

Can we as community or the project people try to reach some of the 96% via targeted e-mail asking them to get back crunching?!

In such an e-mail we can include i.e: tips how to make your 64bit linux machine to work with CPDN; (yeah the TRIFFIDS queue looks insurmountable and the
Linux machines are 59933)

..or to provide other similar tips that can tackle the top few reasons for leaving climateprediction. I know there are studies on crunching behaviour, and some projects still have very large community of crunchers, but one of the most important ones CPDN seems small in numbers.

While e-mail is one option there are probably other options as well.
Care to share ideas? Maybe this should be moved to a new thread first?

A mass email back in 2003/4, to tell a list of people that the project had finally started and they could now sign up, got several of Oxford Uni's email addresses blacklisted. One in America took months to get changed back.
So, no emails.

The main reason for the large number of people no longer crunching is possibly what I heard about back in the 1960s: a study found that for most people, a hobby lasts an average of three years.
And a lot may start out by hearing about one of the BOINC projects and join up, but then find another one that interests them far more.

And there were stories back in 2006/7 on our old BBC project board, from people who joined after the BBC promotion, stayed for a few months, and then found that life was getting in the way, and they had to leave BOINC altogether.

It's also been estimated that 4 or 5 thousand computers would be sufficient to handle the available pool of models. And 10 thousand is about the average size of the largest batch of work that gets released, although very recently, larger batches for specific regional attribution studies are double that. Which still leaves 10,000 as a good user base. (2 models each.)

Having said all of that, Hannah Rowlands, our Communications Officer, has been pushing the project in online media.

I agree. There certainly appears to be enough Windows users. In fact we sometimes experience work droughts that last a week or more.

It might be worth the effort to look into increasing the number of Linux users running this project. At the rate that those Linux only hadam2prm3pm2t are going down they will still be there 2 weeks after the sun burns out.

---

I've thought of another reason why the pool of hadam3prm3pm2t_eu may be decreasing so slowly:
There's a lot of Linux computers without 32 bit libs, and some of these have a LOT of processors, all killing datasets in large quantities.
So of lot of those still showing may be re-issues.

I've thought of another reason why the pool of hadam3prm3pm2t_eu may be decreasing so slowly:
There's a lot of Linux computers without 32 bit libs, and some of these have a LOT of processors, all killing datasets in large quantities.
So of lot of those still showing may be re-issues.

I think that's right: all my current Linux models are reissues - and every wingman a complete disaster ...

and every wingman a complete disaster ...

And that is true of all my Linux wingmen for the short models as well but at least there I have some windows wingmen.

I've thought of another reason why the pool of hadam3prm3pm2t_eu may be decreasing so slowly:
There's a lot of Linux computers without 32 bit libs, and some of these have a LOT of processors, all killing datasets in large quantities.
So of lot of those still showing may be re-issues.

Ok then, can we at least do something about the linux machines? Is there a way to separate reissues*, gone, working and working-but-with-issues and work on the latter? I mean we as community.

I know broken machines are reported when found and blacklisted, but maybe we can try also to reach their owners asking to fix them while providing hints?

CPDN has been around for 10 years, but frankly I haven't heard much about its results**, moreover about its icreased importance after IPCC 5. It is good that now it's been pushed again in the media, but if crunching also raises we may be able to get results faster thus supporting both science and politics.

* I guess this is one machine that is listed several times due to OS change or hardware upgrade?
** I have difficulties grasping the information on the project website (and link it to the models I crunch), as it is not very clear and updated for all models. Something similar was raised at How to find out which project a model belongs to

Is there a way of parsing the stderr output and sending an email to everyone who has a computer which has the missing 32bit lib problem as well as setting the max no of tasks to -1? My impression is that the number of computers involved must be in the thousands so if it has to be done individually for each one and only then when someone reports some of their wingmen it is going to be a very long time till the pr9oblem is solved.

Or of packaging the required library functions in such as way that they can be distributed with the application, rather than making an assumption about what the host will be able to supply in advance?

Yeah I guess I did not mean going after thousands of computers manually.

Can the missing 32libs problem be fixed by BOINC developers including the relevent dependencies in some of the next releases?

Can the missing 32libs problem be fixed by BOINC developers including the relevent dependencies in some of the next releases?

I think the BOINC developers doing anything is unlikely as it is the CPDN executables rather than BOINC that requires them. Hence Richard's idea.

So how should we proceed on the matter?

Not BOINC developers, the BOINC infrastructure provides for different architectures. the 32-bit vs 64-bit ABI dependency is probably compiled in to the core of the core of the core of the HMS Hadley FORTRAN core of the base model.
NOT --

Naah, it's up to the CPDN developers -- there's at least two ways I think of to package the infamous 32-bit libs with the downloads but why bother? 32-bit is so 1990's?("static link, package 32-bit libs) probably many more.

Maybe the project has a clue? What proportion of linux-using-volunteers are limited to 32-bit? 1 in 32? What proportion of supercomputer users are limited to 32-bit math-libs? Huh?
64-bit users -- the vast majority - can compute just as well or better. So why limit the Linux contributors to those who can fight and waste their time with this obsolete 32-bit?

And yeah, this whole thread drift is far from Windows 10 :) Which I hope does well for users of it.

Yeah I guess I did not mean going after thousands of computers manually.

Can the missing 32libs problem be fixed by BOINC developers including the relevent dependencies in some of the next releases?

---

I am still running 32-bit apps on two Linux boxes, including SETI@home, LHC@home, VirtualLHC@home, Einstein@home and CPDN on two Linux boxes with 32-bit OS, alhough one of them has a 64-bit CPU. No additional libraries On my 8.1 Windows PC with a 64-bit AMD A10-6700 I am running also Atlas@home and CPDN 64-bit. But the anz CPDN tasks are extremely slow. I've left them one of 4 cores (or 4 procesisng units, as Windows 8.1 says). I have a 20 GB RAM on that PC, due to Atlas@home needing 4GB RAM each task.
Tullio

---

Hmm, see the OP on Win 10 compatibility as been hijacked somewhat. To come back to the OP, I notice in a recent Grauniad article it seems that the Home version of Win 10 is likely to install updates automatically, not the case with Pro & Enterprise versions.

Recommendation has always been to stop Boinc/CPDN before applying updates, so don't know how this may affect the running of tasks in the future. I normally close everything down before applying update and must admit the update procedure seems to be pretty flawless these days, but it still does need the occasional reboot as part of the process. To me at least, it also seems that the Boinc/CPDN combination is a bit more robust these days and seems to survive things like power loss better than a few years ago.

Does anyone let Win updates happen automatically while Boinc/CPDN is running? I guess only time will tell if the Win 10 Home version update will become an issue.

---

Does anyone let Win updates happen automatically while Boinc/CPDN is running? I guess only time will tell if the Win 10 Home version update will become an issue.

Yes,I do on my Windows 8.1 PC. I have also two projects using Virtual Box, vLHC@home and ATLAS@home on the same PC.
Tullio

---

Not on my two Win7 boxes.

Does anyone let Win updates happen automatically while Boinc/CPDN is running? I guess only time will tell if the Win 10 Home version update will become an issue.

Yes,I do on my Windows 8.1 PC. I have also two projects using Virtual Box, vLHC@home and ATLAS@home on the same PC.
Tullio

Yes, but is anyone running CPDN and letting it automatically update. These models do have a well known tendency to crash if the computer is shutdown without for suspending and then exiting the Boinc manager first. Unexpected reboots could cause lost models.

This could be a problem if there is no way to turn it off.

---

Since I am a beginner Windows user, when I get a Windows update asking me to reboot I simply say yes and reboot. Then I start first BOINC manager, then Virtual Box. All my BOINC programs start regularly.
Tullio

---