1. NXTInfrastructure Avatar NXTInfrastructure
  2. Untitled project
  3. Committee
  4. Issues

NXT Energy Efficiency paper SecondLeo/Matthew C.

Issue #19 new
EvilDave created an issue

https://docs.google.com/document/d/1J8uhdshu9epGRrQHBaloGc4itdvuAHZDAUtNDjOhz-8/edit

Short paper on energy efficiency within the NXT network. To be followed up....

Comments (31)

  2. marcus03

    Here's my feedback on the paper:

    • The paper has more of a marketing tone and not a tone of a scientific/research paper (" beacon of hope", " biggest competitor in the long run", etc.).
    • Missing references for claims. Some that I doubt below.
    • "Curve25519 is the fastest hash out there" I doubt that.
    • "splitting the decryptions between cores is easy to do", but is it actually done with the current NRS implementation. Does it actually use multiple threads and do this threads actually utilize all available cores?
    • Only CPU rates are being looked at. If this would be the only criteria, why isn't the raspberryPi listed?
    • "Assuming the average Nxt transaction is a similar size to Bitcoin". Why assume something that is very easy to find out?
    • Summary: "Nxt is the first coin I’ve seen that actually has a chance of running the coin off of transaction fees less than 1% and are truly significantly lower than the 2%-2.5% that Visa takes." I fail to see how you can get this out of the paper. It doesn't look at existing fees and ROI for forgers at all.

    Despite these and others problems with the paper, I think it deserves funding, if only to motivate the authors to do further work on it. I'd say a funding of 5.000 to 10.000 NXT would be appropiate.

  3. Ian Ravenscroft

    The math in the paper is confusing... It talks about 1250 tps without addressing this on a block rate of 1 minute It uses bitcoin stats to say things are possible rather than using actual NXT stats in the document... It talks about needing 1440 nodes in the network with a flawed justification....

    I do think the author is trying to do something worthwhile - if author can fix all of the above then I agree it would be worth some NXT.

  4. marcus03

    I think we have enough information to decide about this.

    My take: We all seem to see the problems with this paper from a technical/scientific perspective. However, I would like to encourage the authors to further work on this. I'd want to make sure that they receive our feedback and fund them with 5.000 NXT.

  5. Ian Ravenscroft

    Agreed - are we ready to vote on this one?

    5,000 NXT bounty for a paper on the Energy Efficiency on NXT?

  6. marcus03

    Voted yes.

    BTW: This should be announced with a short rationale why we decided like this and mention the feedback we gave on the paper.

    BTW2: We need to wait for rickyjames to officially end the committees vote on Saturday or Sunday.

  9. EvilDave reporter

    From phone, forgive typos. The paper looks impressive at first glance, but is more than a little bit dodgy when u actually look at it closely. It does have possibilities, tho. I'd be tempted to say that we should commit to a 5000 NXT bounty for a revised paper with some of the more dubious assumptions modified.

  10. marcus03

    Seems like we are stuck.

    Here's another approach: Fund the paper with 2500 NXT immediately and another 2500 NXT when it gets a revision based on the feedback we gave.

  13. Ian Ravenscroft

    I'll vote yes but suggest to split 2k now 3k later, but if you want to stick with 50/50 spilt thats ok.

  14. EvilDave reporter

    We have had some input/discussion about the SecondLeo paper from NxtMinnow via PM:

    [quote author=NxtMinnow link=action=profile;u=248810 date=1395083320] Hi EvilDave,

    I just read through the entire Inf Com bitbucket and I think the Committee is doing a stellar job so far.

    I do not think the Secondleo paper can be paid for in good faith. The paper has been presented on the firehose thread, appropriately criticized, and it has showed up in Infcom with no changes having been made from the public criticism.

    The paper has some good information and some bad assumptions. It is not a good read as is and does not accurately reflect Nxt.

    Thanks for your consideration,

    Brian Snyder

    p.s. The total pay I have received for writing about NXT to date is 2900 NXT. Please hold Nxt writers to a peer-reviewed ready to publish standard before handing out bounties. Thank you. [/quote]

    [quote author=EvilDave link=action=profile;u=174118 date=1395096494] Guys:

    OK....I'm including Nxtminnow as a bcc here, for the transparency of it all. I think we should still fund the paper. It has more than a few flaws, but could be a useful contribution to the NXT cause when all the assumptions/dodgy numbers are worked out.

    It certainly represents at least 5-10 hours work (at a minimum), and we should set the precedent of rewarding people for original research/work, if only to encourage others to roll their sleeves up and start contributing. 2500 NXT breaks down to around $100 right now, so it's not like anyones getting rich from this.

    @NxtMinnow: yep, it is kind of unfair. If it helps, I've only recieved about 5000 NXT for all of my work on NXT, the rest of my fairly small NXT holding was bought in just before Xmas 2013. If u want to submit a bounty request for any of your work, I'm prepared to give it some serious attention.

    TLDR: give him 2500, with another 2500 if the paper can be brought up to a much higher standard.

    [/quote]

    [quote author=NxtMinnow link=action=profile;u=248810 date=1395097719] EvilDave and InfCom,

    I think that you are right Nxt should reward for creativity initiative and I back your decision on the 2500/2500 since there is no real application process in place yet. Just like the marketing committee has published their guidelines here: https://forums.nxtcrypto.org/viewtopic.php?f=54&t=875

    I think the same thing should be made clear ala open source principles. If we are ok with paying for a rough, not ready to use (publish) draft today, then ok; for today.

    Just the same as the software must be polished and secure, the writing must be polished as well. Marketing cannot fix punctuation, grammar, and spelling errors. Once the document hits the press, the mistakes stay forever.

    Sincerely, Brian Snyder

    p.s. I even found a spelling error on that massive stack of trifold brochures that none of use caught. "Crypsy" should have been "Cryptsy". Oops, I'm going to have to apologize (probably in person) to Paul Vernon for that one!

    [/quote]

  15. EvilDave reporter

    Yeah, the above post is somewhat untidy, but i'm too lazy to tidy it up any more.

    I think that it's time to finalise this issue, and throw 2500 NXT and our follow-up requirements to SecondLeo.

    Voting.....if there are no other objections, I'll get in touch with SL and klee to arrange release of funds sometime tonight.

  19. EvilDave reporter

    OK, I'm assuming that we are decided on 2k (+3k after revisions)

    I'll organise payment, log it in the wiki and maybe post to InfCom thread starting in about 90 minutes from now, so if there are any last-minute objections , make them quickly.

  20. EvilDave reporter

    I've sent our decision/feedback to secondleo, and a PM to klee to keep him in the loop. Made an entry in the Google spreadsheet for our first payment ....yay! https://docs.google.com/spreadsheet/ccc?key=0AmHnjuFqWSCodGtRMjJ2bVkxeFg4MlBCR2FhNk4tOXc&usp=sharing#gid=3

    Updated the BitB wiki log. https://bitbucket.org/nxtinfrastructure/committee/wiki/Home

    As soon as SL sends me an account number, i'll pass it on to klee for payment of teh initial 2000 NXT.

    Heading over to BTT now to post an update on the InfCom + firehose threads. I'll also open threads on nextcoin and nxtcrypto when/if I get some time.

    Should we keep this BitB issue open, in expectation of a revised version of the paper (my preference) or close it and wait to see if the revised paper shows up?

  22. marcus03

    @EvilDave: Thanks!

    Our spreadsheet is dead. These entries should go to the Wiki. I'll tidy that up.

  23. MatthewC

    First of all, you guys should know that we thought we had this ready to release, and received a bounty of 5000 Nxt already for it before deciding to tidy it up and modify some numbers before release, which is still a work in progress. As you can see some fixes have already been made since someone went through and provided us with feedback. They mostly address chanc3r's comments.

    marcus03, very much appreciate the feedback, will incorporate, responding to your comments:

    -The paper has more of a marketing tone and not a tone of a scientific/research paper (" beacon of hope", " biggest competitor in the long run", etc.). You may be right, you recommend we remove that intro and make this more scientific? Maybe there is something to that suggestion.

    Missing references for claims. Some that I doubt below. -"Curve25519 is the fastest hash out there" I doubt that. Ok, well it's a very fast hash, will re-word. And actually I know for sure it's not the case because Ed25519 is faster, just more complicated

    -"splitting the decryptions between cores is easy to do", but is it actually done with the current NRS implementation. Does it actually use multiple threads and do this threads actually utilize all available cores? Of course it's easy to do, all the verifications do are check is this a valid transaction? And by the time we get to Bitcoin's size I have no doubt this will be implemented. Or does anyone know of something I'm missing?

    -Only CPU rates are being looked at. If this would be the only criteria, why isn't the raspberryPi listed? Good Point, will list.

    -"Assuming the average Nxt transaction is a similar size to Bitcoin". Why assume something that is very easy to find out? I spent a lot of time trying to figure this one out without much luck, SecondLeo found this one last night though.. expected it fixed by the weekend.

    -Summary: "Nxt is the first coin I’ve seen that actually has a chance of running the coin off of transaction fees less than 1% and are truly significantly lower than the 2%-2.5% that Visa takes." I fail to see how you can get this out of the paper. It doesn't look at existing fees and ROI for forgers at all. Whoops, previous version of the paper went a little bit more in depth but we decided to drop that because it was getting too complicated. Will fix that up.

  24. EvilDave reporter

    Just got an account number from SecondLeo for the 2k bounty. I find it a bit out of order that SL and MC didn't tell us about their 5k bounty from Marketing, but I'm still prepared to give them the 2k now.

    They will have to be very charming to get the 3k, though......

    If there are no objections on this, I'll PM klee today to release the funds.

  26. MatthewC

    I told you guys the second I knew you guys were looking at it and considering a larger bounty. Go look at my comment above. Appreciate it though.

    Anyone have any thoughts regarding how we can estimate how many forgers we'll have when we grow to be Bitcoin's size? That's the one last thing holding this paper up at the moment.

  28. EvilDave reporter

    Break it down: How many NXT accounts do we have right now? How many active nodes ? How many are of those are forging?

    Then build it back up based on the number of current BTC users.

    You can only hope to be within an order of magnitude on this sort of estimate. Here's my estimate:

    30,000 NXT accounts currently. (nxtexplorer.com) 350 active nodes (peerexplorer.com) 50 forging (wild guess.com) This gives us an estimate of the forger population between 0.16% of the acc total to 1.16% of the total accounts, which seems way low to me. Now just find out how many active BTC users there are right now.........

    I wish u luck, guys.

  30. MatthewC

    Yeah.. that's a good way to approach it... except every active node is a forging node, right?

    The other issue is that by that point in time, we'll have forging pools implemented and with $100 worth of Nxt in your account, you'll barely ever forge, so those people will still forge, they'll just point their forging power at other accounts, and we won't be able to pay them enough to cover the cost of forging, since we don't have inflation built in, which is why I believe we'll have fewer forgers forging for us. Point is, there is a lot more work involved with estimating that one number than I initially though.. we'll get it done though, I think this is a very important paper for Nxt if we can prove just how much more efficient we are than BTC.

    Yeah, SecondLeo is dealing with a couple of things, so I'm waiting to hear back from him regarding his take on these numbers. We'll get it done soon though. In the meantime back to figuring out how to get the advanced consensus working.. I think I've got it 99% figured out..

    Thanks guys.

  32. Log in to comment
Assignee
–
Type
bug
Priority
major
Status
new
Component
Projects/Bounties to fund
Votes
4
Watchers
5
Jira: the preferred issue tracker for Bitbucket. Join the team!