yeah. That kinda only proves my point.
A lot of trust would be earned if their demands were in the native token instead of the easily fungible, liquid, quick dump in a flash of BTC.
2 Likes
Maybe only for matter of documentation you cannot feed the nossd plot into (unchanged) chiapos prove a challenge or run checks on it. I am not surprised at all this doesnât work.
./ProofOfSpace prove -f ../../nossd/6359A91DD6717DDA2E307F2F94A841B3D12EA476B8D5CECD6DB873A1C14074EA.fpt 793104a9d219394f60b60fd9f8673e1d422ba73e19089a60af59f1dd0a3fab9e
operation: prove
Proving using filename=../../nossd/6359A91DD6717DDA2E307F2F94A841B3D12EA476B8D5CECD6DB873A1C14074EA.fpt challenge=793104a9d219394f60b60fd9f8673e1d422ba73e19089a60af59f1dd0a3fab9e
Caught exception: Invalid plot header magic
terminate called after throwing an instance of 'std::exception'
what(): std::exception
Aborted (core dumped)
./ProofOfSpace check -c 10 -f ../../nossd/6359A91DD6717DDA2E307F2F94A841B3D12EA476B8D5CECD6DB873A1C14074EA.fpt
operation: check
Caught exception: Invalid plot header magic
terminate called after throwing an instance of 'std::exception'
what(): std::exception
Aborted (core dumped)
2 Likes
@Dawson can you improve your query command so it doesnât also start plotting when finished? I want to run millions of challenges against one of your plots.
#i would like to pass a specific file (-f)
./client -f 6359A91DD6717DDA2E307F2F94A841B3D12EA476B8D5CECD6DB873A1C14074EA.fpt --query 793104a9d219394f60b60fd9f8673e1d422ba73e19089a60af59f1dd0a3fab9e
#i would like this command to start and end quickly, without bench marking or starting to plot afterwards
EDIT: there are apparently arguments --no-plotting --no-benchmark ⌠again didnât read the manual, so the only ask remains, make it go faster
2 Likes
@Dawson can you explain the use of the cache file? It appears there is a lock on the file, so 2+ concurrent query calls canât run under the same user
cat /home/nossd/.config/NoSSDChiaPool/cache
2 1 727ccf62e4a572e0dc3a4ef6db0b3b48 1657749732 8a6dd82eeb24bea54ccac1d1186b76de6a72586507d33d3487ec8312f0ee3078
16.170.147.93
1
17144798527676632728 1657749732 1
16.170.147.93
1 Like
The plot format is obviously not the same as the standard Chia format, just feeding it into the Chia binaries expecting plots in the Chia format obviously wonât work.
Have you tried using their binaryâs option to fetch proofs from their plots as described above? I am eager to see the results of analysis by someone on a larger scale, with more proofs, to see if the quantity matches the one expected from a k32. Madmax did it for 10 challenges and verified with chiapos that they are correct.
@xorinox you can use the r (read-only) type in a directory specification -d,r, it wont write anything to this folder then. The cache file contains some minimal amount of data a client may need between runs, the last plotid used for authentication, DNS resolution cache etc, it is also used to protect users from starting two clients accidentally.
Maybe they not just drop the data but put it into the other plots? Query is ony checking a single plot, isnât it?
No. The --query command only returns PoS from one plot. It is intentionally limited this way.
But the command is:
client -d /plots --query
So you canât even choose which plot is used? I was assuming this meant you were challenging all plots, as the returned information shows the plotid it came from. So does it just pick 1 plot at random? Or the first/last one you made?
Can I ask why it was intentionally limited to this? And given what @xorinox says there is no way to test a large amount quickly.
This leaves an unfortunate situation where it is hard to really test this thoroughly.
You might want to think about creating a way for a more thorough test to be performed here, as this is what people need to see to rule out a lot of possibilities.
When doing some extensive test, make sure you only have a single plot is accessible to nossd. Otherwise they could just distribute data across multiple files to fake the space saving.
Yes, basically. It queries one random plot with PoS. It isnât intended to be used on multiple plots and we do not want this command to be used to adapt our client for other pools. One may want to run multiple queries on a single plot because not every query returns a PoS and sometimes there are multiple PoS for one challenge. It is important to ensure that it reads one proof on average.
Speaking about evidence, I already uploaded a complete set of screenshots to the original thread (Chia blocks misplaced/pool activity - went solo, yet block rewards (1.75) transferred to pool - #33 by moranden), yet you did not comment on that at all⌠So please, stop saying this nonsense. This also applies to your repeated reference to threats, which I have no idea where you got it from.
In addition, as mentioned by ojura, your argument with flexpool client is invalid, bcs the issue is related to wallet 
Funny thing is that as soon as someone denies your argument, which is nothing difficult to be honest, you stop communicating.
2 Likes
Ah, I see. Well, the reason you restrict the finding of proofs makes sense from a lock-in point of view for your pool.
But you have given this tool as the main way in which someone should verify that your software does what it claims. You even suggested that Chia Network should verify the claim in the same way.
But being restricted like this, that is simply not a thorough enough verification.
I am not saying that I donât believe any of the claims, by the way. Just that this may create an impasse if a) someone canât test it thoroughly enough, and b) you canât allow more thorough tests without compromising your pool.
Is there anything you can suggest?
We need a way to show that on a full HDD, this software can generate the statistically expected amounts of proofs.
If there is no way to do that then there is no way to verify the base claim.
@Dawson As Iâm testing it as well currently, it would be great if you could add a feature to finalize multiple plots in parallel. Right now I just use sparse plots (for testing ok) because finalizing is the bottleneck as it does it one-by-one in sequential order. As you can mix with -d multiple directories it would be great enhancement.
As @impuls has already noted, testing it on multiple plots will only make the results more questionable. If you have performance concerns, it is something completely separate. --query option was included to test plot validness only.
PS. If you insist on testing on full HDD of plots for whatever reason, then put one plot in a separate folder and run query with a script, but this test doesnât make too much sense.
1 Like
Sorry, but I have to disagree, the âfull HDDâ test with insisting having a bunch of plots on your drive is nonsense. You can do that with a single plot.
You can make a 78.1 (+ a bit more for filesystem overhead) GiB file, mount it as a loopback device, format the device with e.g. BTRFS or EXT4, store the single âcompression level 5â NoSSD plot there and run the proof fetching test inside a docker container a bunch of times. And restrict the NoSSD binary to be only able to read stuff from this filesystem containing the single plot, and no internet access. That way, thereâs no way hidden âup my sleeveâ stuff can happen.
I will be doing this later and posting the results.
If (and only if!) we get on average one proof per challenge for e.g. 1k or 10k challenges, then the Chia community owes these guys a huge apology. These guys may have a big attitude problem, with the entire 100 / 200 BTC demands which came off extremely poorly, but I donât really care about any of that human debris element, Iâm only interested in the tech.
And so far, it looks like that this binary can spit out 101 GiB-K32 equivalent level of proof quantity from a 78.1 GiB plot. I have not seen any evidence to the contrary besides a lot of barking from all sides.
3 Likes
@Dawson You still havenât answered questions about testnet . Have you tried your plots on testnet?
The question about testnet only concerns possible bugs in the pool server code when making an actual block. Thereâs no need to test the plots themselves on testnet.
2 Likes
As long as you can verify that the software doesnât have access to any more space or an internet connection, that would be acceptable, yeah. I suppose the term used should be a full volume, then.
But as I said, I am not being a doubter here. I have leaned more towards believing their claims so far since the first tests verified. However there are still potential holes in what has been tested so far. And as no one has managed to gather enough statistical data yet it still canât be fully concluded that these plots perform as suggested.
So thanks for setting about doing that! I also just want the truth out of this. If I had a spare machine I would be testing this myself even just out of pure curiosity! Although I would say that like many, even if this is proven beyond a doubt to work I wouldnât want to farm with their closed software, anyhow.
But everything NoSSD has done and said so far can easily make sense from their perspective, even if itâs not how many would have conducted themselves, so I am not attacking them in any way.
2 Likes
Agreed. The silence on the testnet question is still confounding. I assume they must have done extensive testing before making it public.