[snip]

Then you'll need very thorough proof that you aren't accusing them of having a fault that you have.

All Boinc does is schedule, it isn't doing the science. There are many far more complex programs out there, written sometimes by one single programmer, on the freeware market. And they work.

And it's not the programming per se that sux, it's the decisions on how it should operate.

Still, you'll need to prove that you can do good programing work

Probably including releasing it as freeware, so many other programmers get a chance to give their opinions on how good or bad it is.

It's less important whether your programming work is for scheduling or for science, as long as you can convince enough people that it does something useful that isn't already available elsewhere.

[snip]

Then you'll need very thorough proof that you aren't accusing them of having a fault that you have.

All Boinc does is schedule, it isn't doing the science. There are many far more complex programs out there, written sometimes by one single programmer, on the freeware market. And they work.

And it's not the programming per se that sux, it's the decisions on how it should operate.

Still, you'll need to prove that you can do good programing work

Probably including releasing it as freeware, so many other programmers get a chance to give their opinions on how good or bad it is.

It's less important whether your programming work is for scheduling or for science, as long as you can convince enough people that it does something useful that isn't already available elsewhere.

Or is better or is just as good but cheaper than the other guy.

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

---

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

When are you going to prove that you can do a better job?

If it's for science, make sure that it's closer to qualifying for a Nobel prize than an Ig Nobel prize.

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

And from where do you propose they get the funding - are you going to pay them?

The problem with BOINC is all the legacy code that is still in the source. Too many patches upon patches. Half the code does nothing now as the hardware and project types have moved on from what BOINC was originally written for. BOINC needs a complete fresh slate rewrite.

And BOINC is still trying to be all things for every platform and every user. Will never be optimized for any platform. Should have separate clients for low power devices like phones, cpu only hosts, gpu only hosts etc. etc.

And if you are a capable developer, you can make changes to the source to improve the client and just compile it for your own specific needs. Just don't bother submitting it to the official BOINC developers because it will never get out of committee and will languish in the proposed tree forever.

And BOINC is still trying to be all things for every platform and every user. Will never be optimized for any platform. Should have separate clients for low power devices like phones, cpu only hosts, gpu only hosts etc. etc.

BOINC is just a manager- it manages computing resources. and the applications that make use of them. So you only need the one manager, just different application versions to run on different hardware/operating systems.

The biggest issue is that BOINC was developed when there were just CPUs (single core/thread CPUs at that) with a few dual and quad socket systems. It has evolved to work with GPUs as well as CPUs, and the present system of tracking resources by application helps it deal with the differences in compute capability. It is up to the application developers to develop their applications to make appropriate use & take in to considerations the limitations of the resource they intend to use (ie phone, tablet, conventional computer, GPU, Quantum computer etc) and then BOINC determines how to use those applications to process work to meet the settings the user has chosen.

I agree BOINC needs to be re-done from scratch, making use of all the lessons of the last 20 years- in particular the server side scheduler code. Most importantly it needs to be compute resource agnostic, and when a single system has multiple compute resources (eg a computer that has an x86 CPU, arm CPU, and a GPU), allocation of work needs to be done by the manager, not the server (as it is now) so, for example- if necessary the manager can transfer work between CPU/GPU/whatever to keep those resources busy if there is a shortage of new work, according to the users project preferences.


But the simple fact is that even if fully rebuilt from scratch, all the work will be based on the needs of the vast majority of users. Those that want to do things differently from what BOINC was designed to do (allow different compute resources to be used to do work for different projects in proportion to people's preferences) will still be able to do so, using configuration files. The time and effort required to cover every possible configuration desire using a GUI would be never ending, and a waste of time & effort with no end point.

---

Grant
Darwin NT

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

When are you going to prove that you can do a better job?

If it's for science, make sure that it's closer to qualifying for a Nobel prize than an Ig Nobel prize.

I can do a better job at driving a bus than all the bus drivers around here, doesn't mean I'm going to.

---

And BOINC is still trying to be all things for every platform and every user. Will never be optimized for any platform. Should have separate clients for low power devices like phones, cpu only hosts, gpu only hosts etc. etc.

BOINC is just a manager- it manages computing resources. and the applications that make use of them. So you only need the one manager, just different application versions to run on different hardware/operating systems.

The biggest issue is that BOINC was developed when there were just CPUs (single core/thread CPUs at that) with a few dual and quad socket systems. It has evolved to work with GPUs as well as CPUs, and the present system of tracking resources by application helps it deal with the differences in compute capability. It is up to the application developers to develop their applications to make appropriate use & take in to considerations the limitations of the resource they intend to use (ie phone, tablet, conventional computer, GPU, Quantum computer etc) and then BOINC determines how to use those applications to process work to meet the settings the user has chosen.

I agree BOINC needs to be re-done from scratch, making use of all the lessons of the last 20 years- in particular the server side scheduler code. Most importantly it needs to be compute resource agnostic, and when a single system has multiple compute resources (eg a computer that has an x86 CPU, arm CPU, and a GPU), allocation of work needs to be done by the manager, not the server (as it is now) so, for example- if necessary the manager can transfer work between CPU/GPU/whatever to keep those resources busy if there is a shortage of new work, according to the users project preferences.


But the simple fact is that even if fully rebuilt from scratch, all the work will be based on the needs of the vast majority of users. Those that want to do things differently from what BOINC was designed to do (allow different compute resources to be used to do work for different projects in proportion to people's preferences) will still be able to do so, using configuration files. The time and effort required to cover every possible configuration desire using a GUI would be never ending, and a waste of time & effort with no end point.

It needs to be rewritten from scratch and vastly simplified. At the moment Boinc can never decide what to do because it can't handle preferences set in config files, in my preferences on the project, and even more hidden ones the server works out. Also basic simple stuff needs an injection of common sense. Here's a simple example:

Let's say you have an 8 core CPU. You join two projects, X at weight 4, and Y at weight 1. Project X produces 6 core tasks. Project Y produces 1 core tasks. The only possible sensible thing to run in this case is a 6 core task from project X, and 2 single core tasks from project Y. That fully utilises the CPU and is the closest to the user's weighting request. But Boinc can't even do that. Because that would produce a weighting of 3 to 1, but the user asked for 4 to 1, which is a higher ratio. Boinc wants to do more of the 6 core tasks. But it can't fit two 6 core tasks into the CPU, so it leaves two cores idle!! How monumentally ridiculous.

---

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

And from where do you propose they get the funding - are you going to pay them?

All these projects have loads of money. Do you think the Large Hadron Collider was built out of scrap parts from freecycle?

---

But it can't fit two 6 core tasks into the CPU, so it leaves two cores idle!! How monumentally ridiculous.

Yeah, doing what the user tells it to do really isn't what software should be doing.

---

Grant
Darwin NT

But it can't fit two 6 core tasks into the CPU, so it leaves two cores idle!! How monumentally ridiculous.

Yeah, doing what the user tells it to do really isn't what software should be doing.

You can't be that dim. The weighting given by the user is how important each task is, and not an absolute requirement. Clearly the user will want his computer to be doing something useful, even if it's not the most important thing (just as happens if the important project is out of work).

If I tell you that mowing the lawn is very important, and weeding the flowerbeds is of far lesser importance, and you have three workers and two lawnmowers, what would you assign them to do? Boinc would put one of them on a bench.

---

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

And from where do you propose they get the funding - are you going to pay them?

All these projects have loads of money. Do you think the Large Hadron Collider was built out of scrap parts from freecycle?

The problem with the Projects paying for the development is each wants ITS Project given priority over EVERYTHING!! ie no no no that hurts my algorithym and that's unfair and I'm taking my money and going home if you do that!!

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

And from where do you propose they get the funding - are you going to pay them?

All these projects have loads of money. Do you think the Large Hadron Collider was built out of scrap parts from freecycle?

The problem with the Projects paying for the development is each wants ITS Project given priority over EVERYTHING!! ie no no no that hurts my algorithym and that's unfair and I'm taking my money and going home if you do that!!

Then Boinc should charge a small fee to each project to use its services, employ decent programmers, and not cater to any one project over another. But if say 5 of the projects thought of something that would help them all out, then do that.

---

[snip]

And from where do you propose they get the funding - are you going to pay them?

All these projects have loads of money. Do you think the Large Hadron Collider was built out of scrap parts from freecycle?

All of the BOINC projects I've found mentioning how much money they had were asking for donations, not saying that they had loads of money.

If you insist that many of them have loads of money, you'd better supply proof of that.

One that might is Rosetta@home, due to the large license fees they charge for using their software on sites other than theirs (and their test site, Ralph@home).

[snip]

The problem with the Projects paying for the development is each wants ITS Project given priority over EVERYTHING!! ie no no no that hurts my algorithym and that's unfair and I'm taking my money and going home if you do that!!

Then Boinc should charge a small fee to each project to use its services, employ decent programmers, and not cater to any one project over another. But if say 5 of the projects thought of something that would help them all out, then do that.

If that's what you want, find a source of money to pay for programmers with enough experience that you'd call them decent, instead of the university students that most of them are now.

Or is better or is just as good but cheaper than the other guy.

And here lies the problem. Boinc uses free workers. Pay peanuts, get monkeys.

And from where do you propose they get the funding - are you going to pay them?

All these projects have loads of money. Do you think the Large Hadron Collider was built out of scrap parts from freecycle?

The problem with the Projects paying for the development is each wants ITS Project given priority over EVERYTHING!! ie no no no that hurts my algorithym and that's unfair and I'm taking my money and going home if you do that!!

Then Boinc should charge a small fee to each project to use its services, employ decent programmers, and not cater to any one project over another. But if say 5 of the projects thought of something that would help them all out, then do that.

But that's NOT the Boinc model first written up by Dr A way back when. Boinc is open source to anyone and everyone and there are several places Admins can go to get help if they need to do so. We users normally get help thru the Message Boards at each project.

You can't be that dim.

Try laying off the insults & stick to the topic. Then people might take you and any suggestions you make more seriously.

The weighting given by the user is how important each task is, and not an absolute requirement. Clearly the user will want his computer to be doing something useful, even if it's not the most important thing (just as happens if the important project is out of work).

And for other projects that's exactly what does happen. Of course you are using an example where the project's application and the computing resources makes doing so impossible. With all other applications that don't need multiple cores/threads for a single application, there will be no unused cores/threads.

And once again i will point out the very basic fact- resource share is something that is honoured over time. At any given time, work done won't necessarily match the resource share settings due to application requirements, hardware being used & the users settings. As long as you continue to ignore that fact, you will forever be disappointed with what may be occurring at any given time on your system.

If I tell you that mowing the lawn is very important, and weeding the flowerbeds is of far lesser importance, and you have three workers and two lawnmowers, what would you assign them to do? Boinc would put one of them on a bench.

As would the gardening company if it were a requirement that there be a minimum of 2 workers performing any given task- which is the situation in the example you quoted.

If you actually have a valid complaint, people would take you seriously. But if you complain about things being done according the the users requirements, as limited by their hardware & the applications being run, you have nothing to complain about.

---

Grant
Darwin NT

All of the BOINC projects I've found mentioning how much money they had were asking for donations, not saying that they had loads of money.

How do you think they pay their staff, pay the electricity bill, etc? A bit of programming shared between many projects would be chicken feed compared to that.

If you insist that many of them have loads of money, you'd better supply proof of that.

Or what? You'll come round my house and duff me up?

One that might is Rosetta@home, due to the large license fees they charge for using their software on sites other than theirs (and their test site, Ralph@home).

And WCG, sponsored by IBM.

---

If that's what you want, find a source of money to pay for programmers with enough experience that you'd call them decent, instead of the university students that most of them are now.

You'd only need a small number for a short time.

---