On Mon, Dec 28, 2015 at 2:12 PM, Peter Todd via bitcoin-dev <bitcoin-dev@lists.linuxfoundation.org> wrote:

Do you specifically mean selfish mining as defined in Emin Gün
Sirer/Ittay Eyal's paper? Keep in mind that attack is only a significant
issue in a scenario - one malicious miner with >30% hashing power -
where you're already very close to the margins anyway; the difference
between a 50% attack threshold and a 30% attack threshold isn't very
significant.

This is not quite right: we know that selfish mining is a guaranteed win

at 34%. We do not know when exactly it begins to pay off. The more 

consolidated and centralized the other mining pools, the less of a threat

it is below 34%; the more decentralized, the more likely it is to pay off 

at lower thresholds.

Far more concerning is network propagation effects between large and
small miners.

On a related note, the Bitcoin-NG paper took a big step towards moving

these kinds of concerns out of the realm of gut-feelings and wavy hands 

into science. In particular, it introduced metrics for fairness (i.e. differential

rate in orphans experienced by small and large miners), hash power 

efficiency, as well as consensus delay. 

 

For that class of issues, if you are in an environemnt
where selfish mining is possible - a fairly flat, easily DoS/sybil
attacked network topology - the profitability difference between small
and large miners even *without* attacks going on is a hugely worrying
problem.

Indeed, there is a slight, quantifiable benefit to larger pools. Which is why

we need to be diligent about not letting pools get too big.

 

Note though that Eligius is *not* the only pool to have had problems

with block withholding, though AFAIK Eligius is the only one who has
gone on record so far. (as I said in my original post, I'm relaying
information given to me under condition of confidentiality)

I can see why they don't want to go public with this: it means that they

are less profitable than other pools.