Reward token mechanism is completely broken - impossible to truly 'leave' theDAO once the first proposal has passed

[eththrowa]

Right now a new proposal framework has been submitted that is supposed to fix some of the most pressing problems with the DAO - it tightens up the rules around withdrawals and throws exceptions if a bunch of extra conditions are not met.
It claims that this solves the dilution of reward token issue by making it the job of the curators to pick only real proposals - but ignores that this flaw is a result of the reward token issuance being a completely broken mechanism which does not perform as it should. This means that it is effectively impossible to leave theDAO even if you successfully split. Here I will explain why.
For those of you who haven't read through the contract code; The way it works is that every time a payment is made as the result of a successful proposal reward tokens equal to the amount of wei paid are issued. with 1 wei = to 1 reward token. Reward tokens are 'held' by either theDAO or they are transferred when someone splits off. Ownership of reward tokens is how DTH entitlement to any reward payment received by theDAO is calculated.
The main problem is that reward tokens are not proposal specific - the logic is simply: the more money theDAO spends the more reward tokens get produced. This is the cause of the 'dilution' attack - an ability of the parentDAO to reduce the value of the reward tokens held by child DAO's by approving proposals for large sums which generate more reward tokens but feed the money back into theDAO.
But the dilution attack is just on aspect of how the reward token system doesn't work and why leaving theDAO doesn't mean what you think it does.
Issue 1: Never leave me!
You are a DTH and are pretty stoked that the first proposal to theDAO has just passed; slockit has just received 2 million ether and you couldn't be happier about it. But then dumbcorp shows up and asks for 3 million ether to club baby seals - it looks like its going to pass and you want out of theDAO. So you quickly split and take your share of the reward tokens (5%) to your new DAO - this means that you should receive 5% of the rewards produced by slockit from now until forever right? WRONG!! Your reward tokens entitle you to a share of all the proceeds of every proposal ever approved by theDAO even if you leave. Because dumbcorp's proposal passed and they wasted all their money - you now only get 2% (20000000.05/5000000) of the reward sent to theDAO by slock.it. This effectively means you will in part finance every single proposal that theDAO green lights. *Every. Single. One.** Meaning you just clubbed some seals.
Issue 2: The price is wrong!
You are a DTH and theDAO has just funded slock.it 2 million ether (approx 30m USD), when something amazing happens - the price of ether goes up 10x!! Everyone is pretty excited by this news and a lot of people exit the DAO (50% split off). theDAO still has plenty of cash though so it can keep on funding projects except now we have a problem... Because of the way reward tokens are calculated any rewards generated by subsequent proposals will mostly go to people who invested in slock.it and left. Reward tokens are calculated by the number of wei spent in proposals - if there is a dramatic (10x) upward change in the dollar value of ether a proposal that costs theDAO 30m dollars afterward will give those left in theDAO less than 10% of the reward tokens that were generated for the slock.it proposal. This means that those individuals who left theDAO are going to receive close to 45% of any rewards generated by this second proposal - despite not voting for it or 'contributing' to it.
There are all sorts of other weird and distorted outcomes that this mechanism produces when people leave or the price changes and it needs to be fixed before any proposals pay out ether. The long and short of it is that generic reward tokens need to be dropped entirely and replaced with unique tokens generated for each proposal - this is the only way to fix these issues.
TLDR: When you take part in a proposal you are not buying into the proposal, you are buying into theDAO and all its subsequent decisions - and the mechanism for doing this is so broken it is incredibly unlikely that you will get your 'fair share'.

[thecryptofiend]

Interesting. I thought that each project had unique reward tokens but if not then it is certainly problematic.

[Hibryda]

As it gets interesting, what other scenarios with split are possible?
Could you point out to places in code?

[North]

The reward tokens are proposal specific. And if you split from the DAO and stop funding the advancement of a particular proposal, there is an opportunity that your reward tokens, that you funded before the split, could be diluted by future proposals that your split DAO is not funding.

The Github guide also helps explain how rewards work:

https://github.com/slockit/DAO/wiki/Understanding-the-DAO-accounting

[eththrowa]

They aren't proposal specific - they are generic and generated whenever money gets paid out. Here is the relevant contract code:

// Tracks the addresses that own Reward Tokens. Those addresses can only be
// DAOs that have split from the original DAO. Conceptually, Reward Tokens
// represent the proportion of the rewards that the DAO has the right to
// receive. These Reward Tokens are generated when the DAO spends ether.
mapping (address => uint) public rewardToken;
// Total supply of rewardToken
uint public totalRewardToken;

[Hibryda]

One rationale behind common actions is to have a volume. Splitting generates risk of diluting funds pretty quickly. Because people have different ideas and everyone thinks his/her idea is the best.
As many tokens are just held for speculation the tyranny of majority could actually work. Provided that people would take rational and responsible decisions in organized way. Which they didn't.
But as to reward tokens and this bug that OP shown in separate topic - this is an issue that needs a fix. And the best solution is to have proposal specific rewards. At the moment of course.

[thewalkingdad]

I'm surprised that this is surprising to so many...

An approved and passed proposal is binding for all DTH. This is just another reason why voting (with all of its current problems) is still important.

There are definitely some critical flaws with the way this DAO has been established. I look forward to voting YES to smooth some of these areas out.

[Hibryda]

Surprising is not that there are flaws, some of them pretty serious. What is surprising is the fact that this initial experiment got out of control.

[TheDaoist]

I'm not going to pretend to understand what anyone is talking about here because I am lost. But I think I get the gist. The solution to this problem is since it is code based, write a new code solution that fixes it.

Right now my only understanding is that the flawed mechanisms for how theDAO works is being addressed at lengths and that corrections they make to fix the other issues may fix this by proxy (or not), who this they is I don't know, they may just be waiting for members of the community to code something as a solution.

In which case, Reward Tokens being proposal specific, is not among that list yet. If not being proposal-specific is truly important issue to be addressed then code a solution and write a proposal for what that code does in English.

TLDR: if it's a problem, someone should propose a way to fix it and offer a smarter, alternative code that makes it proposal-specific and addresses the problems you have mentioned here.

[Beez]

Proposal-specific reward tokens proved to be very complicated and difficult to do, so it wasn't implemented for 1.0. I'd like to get it working for 2.0, and I have a theory, but it needs to be tested which requires time on my part.

Actually, if anyone has a link to a 2.0 development thread, I'd like to put the theory there so others can work on it and see if it's useable.

[thewalkingdad]

Not explicitly 2.0, but is focused on bettering our DAO community.

The Reconstruction of the DAO: Solutions for All Problems

Current Problems Most DAO token holders (DTHs) are jointly liable for all business activities of the DAO. DTHs are also individually liable for purposes such as taxation, business registration, and legally trading DAO tokens. The governance structure of the DAO is not flexible - for example, if a shareholder is willing to stake $10 of his own assets for a $10 proposal, but 51% voted no he is unable to back this proposal. The inverse is true when a majority vote dictates the use of the minorities funds. The DAO wallet address currently holds $175 million worth of ether which is not being utilized. In a 1% savings account this amounts to $5,000 in lost interest a day. This discussion includes various solutions that come together to form an answer to all of these problems and promote …

[macks]

Actually I like the way the dao is designed right now.. To have some incentive to not have it split into little pieces too fast. Lets be and stay a decentralized entity that somehow forms an identity known to others and not confuse everyone with dozens of litle dao parts...
Also your first point critizises that there's too much incentive to stay with the dao and you second point somehow critizises that there's not enough incentive to stay with the dao....
How do you imagine all of this should be calculated? lets say slockit gets fundet in multiple proposals but has only one product giving a fixed percentage of its profit to the dao.. how do you want to calculate how much of these profits go to which proposal..?
Also some proposals can just fund the overall development of ethereum and help the eco system.. It makes sense how the dao is designed right now in that regard..