GPU72's fetching/reporting API

[Aramis Wyler]

Originally Posted by swl551

Here are my thoughts on how the repository, application and user can interact over the API.

Up side to re-doing the API: Better api! More familiar to modern web app devs!
Down side to re-doing API: Prime95 will cease to function wrt gpu72.

My take on this is that we need to build the gpu72 api so that it is indistinguishable from the primenet api by prime95. But there is no reason we can't also do a more modern web service style api with SOAP and RESTful states. Those could simply be translated to primenet style strings by gpu72 and passed on or consumed w/o the remote apps being the wiser. It could even function as an assignment fetcher for non-gpu72 work if that were dubbed reasonable.

[swl551]

Originally Posted by Mini-Geek

As described, it would be trivial for someone to spoof an app GUID (even if encrypted en route, it can be decrypted at the client end). If this is intended as a security mechanism, this is not a suitable one. If merely as a polite way to say what app is what, then it's fine.

yes, a polite way.

[chalsall]

OK, I'm just back from a *really* long and hilly hash, and have 22 minutes to turn around to leave for dinner (I'll take five minutes; my girlfriend and her cousin -- longer)...

However, one thing I think we should keep in mind here is we shouldn't overstep our mandate. In my mind, the API we're talking about is the one GPU72 will expose to clients to fetch TF (and possibly P-1) work, report the status of assignments, etc.

We should not presume that George, Scott (and/or possibly James) will ever change the existing Primenet API from a design which already exists, has been working for years, and has tens (possibly hundreds) of thousands of deployed clients using.

I will speak to this in more depth tomorrow.

[swl551]

So determining backward/forward compatibility is knowing what applications will use gpu72/api.

I assumed it was Spidey, Fetcher, MISFIT. The thought of having thousands of prime95.exe instances calling gpu72/api never occurred to me. (I'm not sure you stated that was a condition, but I never thought about it)

[chalsall]

Originally Posted by swl551

So determining backward/forward compatibility is knowing what applications will use gpu72/api.

I assumed it was Spidey, Fetcher, MISFIT. The thought of having thousands of prime95.exe instances calling gpu72/api never occurred to me.

[Still waiting for my girlfriend and her stupid cousin to be "presentable"... ]

One thing to keep in mind is a couple of tricks I've had to use to be able to allow Prime95/mprime clients to use the Proxy (slightly different than the (being discussed) GPU72 API) for assignments.

1. To allow LL/DC candidates to be truly transferred to the owner's account, the candidate held by GPU72 for assignment must be owned by Anonymous.

2. GPU72 cannot alter any message passed onto Primenet. It must either be exactly the same (truly proxied), or else intercepted and fully answered by GPU72. This is why P-1 update reports don't show up on the Primenet users' reports.

Originally Posted by swl551

(I'm not sure you stated that was a condition, but I never thought about it)

I didn't. But I just wanted to be absolutely clear. I got the impression that Bill might have been thinking along those lines (although I might have misread).

[Dubslow]

Originally Posted by chalsall

I didn't. But I just wanted to be absolutely clear. I got the impression that Bill might have been thinking along those lines (although I might have misread).

I got confused there for a second.

[Aramis Wyler]

Originally Posted by Aramis Wyler

Down side to re-doing API: Prime95 will cease to function wrt gpu72.

No, I knew there was no way we were changing the api. That doesn't rule out having a second api though, if Scott wanted to take the time to build it.

[Aramis Wyler]

• Originally Posted by chalsall

  GPU72 cannot alter any message passed onto Primenet. It must either be exactly the same (truly proxied), or else intercepted and fully answered by GPU72. This is why P-1 update reports don't show up on the Primenet users' report.

That is important to know, though it doesn't end up being much of an impedement unless we wanted to make it one. For example, say someone (using misfit or fetcher or prime95, whatever) uses the gpu72 api to request a number w/an exponent > 65m to p-1 and we don't have one. Rather than return a failure, we could pass it on to primenet after we've failed to fulfill it ourselves. If prime95 has something that meets the need we log the result and pass it back to the client. If someome sent a request to unreserve a number for whatever kind of work we simply check who assigned it - gpu72 or prime95 and either deal with it or pass the request on. I'm not 100% on how Anon spidey reserves numbers from primenet, but I assume it doesn't use the api. I assume that because you don't seem to have the secret hash and if anon spidey had used the api to reserve a number anonymously, wouldn't primenet have assigned it the secret hash and poof problem solved?

Altering the request might have been a hair faster than dealing with it or making new requests/queries and sending them out to primenet on anonspidey's behalf, then transfering control of the resulting number to the originator, but it is not insurmountable if that restriction is in place.

Here's another sort of thing that I think we might not have to deal with:

Originally Posted by chalsall

(Ironically, the core code has been working for months. The hold up is I have to add some new functionality to the web UI in order to allow people to create new GPU "Machines" on the site, and get the "secret". "Life" (and a death in the family) got in the way, but things are beginning to calm down again....)

I ask myself why. Why would we bother creating machines on this site? Primenet already implemented that functionality, and if we note the users request for a new machine and pass it on to primenet, note the response (success or not) log the new machine (or not) complete with primenet approved guid, and send the response back to the client. In a lot of these cases, stuff done on the web site would only be the gpu72 website talking to it's own api as a client but I think creating a new machine might be one than needs to come from the machine via an actual client like misfit or prime95. Why reinvent the wheel?

[NBtarheel_33]

Question: Why not combine the best attributes of PrimeNet and GPUto72, set up collaboration with George and Scott K., and plan on replacing PrimeNet v5 with a fusion of PrimeNet, MISFIT, GPUto72, etc.? We need a new server, anyway. Might as well take the opportunity to get everything seamlessly integrated and under one umbrella.

[Dubslow]

Originally Posted by NBtarheel_33

Question: Why not combine the best attributes of PrimeNet and GPUto72, set up collaboration with George and Scott K., and plan on replacing PrimeNet v5 with a fusion of PrimeNet, MISFIT, GPUto72, etc.? We need a new server, anyway. Might as well take the opportunity to get everything seamlessly integrated and under one umbrella.

Inertia (sadly -- it's not like I personally can do much about it).


(Perhaps if some fundraising was undertaken, we could redo PrimeNet and put chalsall in what has been Mr. Kurowski's position?)

[chalsall]

Originally Posted by NBtarheel_33

Question: Why not combine the best attributes of PrimeNet and GPUto72, set up collaboration with George and Scott K., and plan on replacing PrimeNet v5 with a fusion of PrimeNet, MISFIT, GPUto72, etc.? We need a new server, anyway. Might as well take the opportunity to get everything seamlessly integrated and under one umbrella.

I would welcome that opportunity. Both James and I have offered to help with Primenet V6 for some time. I understand James is continuing to wrap is head around the current Primenet code base (he has access to it).

Originally Posted by Dubslow

(Perhaps if some fundraising was undertaken, we could redo PrimeNet and put chalsall in what has been Mr. Kurowski's position?)

That would be very inappropriate. Scott K. was involved from the beginning -- I'm a relative newcomer.

I do think, however, that something like a "Primenet Development Committee" being created would be a good thing for GIMPS.