I see posts in this forum on optimized project apps, but not on optimized BOINC clients. I found this link, which offers optimized BOINC clients with and without "calibration," which apparently keeps you from getting inflated credit per WU. So what's the deal? Are the optimized clients just for generating additional credit, or is there some productivity benefit that actually gets more crunching done per hour? Is there a reason this isn't adopted as standard by the BOINC community as a whole? Personally I'm not in this to be claiming credit that's not rightfully mine, so if it's just a credit thing, it seems a bit ill-conceived IMHO. But if more science is getting done, isn't that a good thing? Or is this just a way to scam the system? Or am I missing something?

Usually there is no additional benefit any more, it's plain cheating.

In the beginning of optimisation @SETI (and only there, never in any other project) there was a need for the calibration of the benchmarks if you wanted to claim the same amount of credits for your optimised application as everyone else was claiming while running stock app.
Since the transition to fixed credits this is no longer necessary.
In every other project this was always and ever plain cheating as they never had the needed corresponding optimised applications to justify the artificial benchmark inflation.

Another reasonable reason for some to fiddle with the benches was the poor benchs on linux systems, but with the correction of that since 5.8.8 there is no need for Linuxers any more as well.

So the only reason for the use of the so-called optimised client (as it doesn't do anything sciencewise there is no need for optimisation, it doesn't benefit the projects anyhow) is to claim too much credits, imho the use of them should be discouraged and , if possible, forbidden. Crunchers still running 5.5.0 and other cheater clients should imho get no credits at all in projects still using benches as a punishment for their unfairness.

Best to get rid of those unfairness would be fixed credits, but it seems not too easy to implement in all projects.

Not completely true. There are two valid reasons I know of to sun custom complied BOINC clients.

1) CPU affinity. It can improve performance slightly, for multi-core machines where the L2 cache is not shared. Examples are the Pentium D, or any multi-chip machine.

2) Return results immediately. This has nothing to do with points, but it provides functionality that some desire.

And project admins don't.

I'm not making judgement calls here. Just explaining the functionality. Right/Wrong is for each project to define.

If a client can have return-results-immediately, how would the project decide if it wants that functionality or not?

I not follow your question. They make that decision like any other decision. However, if they decide that they don't want it, there is nothing I know of to enforce. Perhaps they could do things like allowing clients to connect no more frequently than every x minutes.

The projects don't have any say over the client, it's on your computer, it's you who is responsible for it. If you let him contact the server every 10 minutes, they cannot do anything against it.

Okay. So what's your point?

Return results immediately is not good for the projects, it should not be done.
But...
It's not possible for the projects to regulate it, as it's only done on the client side, so the project cannot decide anything about it. It's up to the user to behave in a way the projects want, do behave decently, without "return results immediately".

I agree. If the projects don't want it used, the members should do it.

But again, I never said it was right or wrong. I was explaining that there are reasons for optimized clients, beyond "plain cheating".

One question:
What's the use for this feature?

It's unwanted by the projects and has no benefits I can think of.

It is a feature that can be needed and used during project development. That way the the project can get results back quickly without having to wait, or manually click "Update" on the BOINC clients.

Getting low latency is really something that should be defined via parameters on the server, not by requesting people to use clients with discouraged features. I don't think the real purpose of RRI was for the benefit of low latency projects anyway.

I think you misunderstand what I was saying. Saenger asked the purpose of the feature. It is generally used for developers, so no waiting is required to get back results. Not for users like you and me, to achieve "low latency".

Well, if its really used in that fashion, I haven't seen it.

Probably not. This would be inside the labs, as they are developing and testing the applications. Before they are released into the wild.