Makes sense.. So with that said, I'd propose the following criteria for selecting UTXOs:

1. Select the smallest possible set of addresses that can be linked in order to come up with enough BTC to send to the payee.
2. Given multiple possible sets, select the one that has the largest number of UTXOs.
3. Given multiple possible sets, choose the one that contains the largest amount of total BTC.
4. Given multiple possible sets, select the one that destroys the most bitcoin days.
5. If there's still multiple possible sets, just choose one at random.

Once the final set of addresses has been identified, use ALL UTXOs from that set, sending appropriate outputs to the recipient(s), a new change address, and a mining fee.

Miners should be cognisant of and reward the fact that the user is making an effort to consolidate UTXOs. They can easily spot these transactions by looking at whether all possible UTXOs from each input addresses have been used. Since most miners use Bitcoin Core, and its defaults, this test can be built into Bitcoin Core's logic for determining which transactions to include when mining a block.

--
James G. Phillips IV  
"Don't bunt. Aim out of the ball park. Aim for the company of immortals." -- David Ogilvy

 This message was created with 100% recycled electrons. Please think twice before printing.

On Sat, May 9, 2015 at 3:38 PM, Pieter Wuille <pieter.wuille@gmail.com> wrote:

Miners do not care about the age of a UTXO entry, apart for two exceptions. It is also economically irrelevant.
* There is a free transaction policy, which sets a small portion of block space aside for transactions which do not pay sufficient fee. This is mostly an altruistic way of encouraging Bitcoin adoption. As a DoS prevention mechanism, there is a requirement that these free transactions are of sufficient priority (computed as BTC-days-destroyed per byte), essentially requiring these transactions to consume another scarce resource, even if not money.
* Coinbase transaction outputs can, as a consensus rule, only be spent after 100 confirmations. This is to prevent random reorganisations from invalidating transactions that spend young coinbase transactions (which can't move to the new chain). In addition, wallets also select more confirmed outputs first to consume, for the same reason.

On May 9, 2015 1:20 PM, "Raystonn" <raystonn@hotmail.com> wrote:

That policy is included in Bitcoin Core.  Miners use it because it is the default.  The policy was likely intended to help real transactions get through in the face of spam.  But it favors those with more bitcoin, as the priority is determined by amount spent multiplied by age of UTXOs.  At the very least the amount spent should be removed as a factor, or fees are unlikely to ever be paid by those who can afford them.  We can reassess the role age plays later.  One change at a time is better.

On 9 May 2015 12:52 pm, Jim Phillips <jim@ergophobia.org> wrote:
On Sat, May 9, 2015 at 2:43 PM, Raystonn <raystonn@hotmail.com> wrote:
How about this as a happy medium default policy: Rather than select UTXOs based solely on age and limiting the size of the transaction, we select as many UTXOs as possible from as few addresses as possible, prioritizing which addresses to use based on the number of UTXOs it contains (more being preferable) and how old those UTXOs are (in order to reduce the fee)? 

If selecting older UTXOs gives higher priority for a lesser (or at least not greater) fee, that is an incentive for a rational user to use the older UTXOs.  Such policy needs to be defended or removed.  It doesn't support privacy or a reduction in UTXOs.

Before starting this thread, I had completely forgotten that age was even a factor in determining which UTXOs to use. Frankly, I can't think of any reason why miners care how old a particular UTXO is when determining what fees to charge. I'm sure there is one, I just don't know what it is. I just tossed it in there as homage to Andreas who pointed out to me that it was still part of the selection criteria.