The attacks yield zero profit, and in actuality cost money. Only a government agency (or something similar) not driven by profit and highly motivated to stop bitcoin would bother with these types of blockchain/UTXO bloat attacks.

Doesn't make sense that the protocol allows non standard TX's to be accepted in block form by nodes yet at the same time they won't relay or accept them if simply propagated.

The transactions were always valid.

To make them invalid would require a hard fork. We try to avoid hard forks for obvious solutions.

So the compromise was to change bitcoind nodes to only consider certain types of transactions as "standard" and not re-transmit transactions that did not qualify as "standard", these have since been called "non-standard". Along with this change many miners decided to not include these "non-standard" transactions either (but some still do).

This did not require a fork because the base protocol, which is defined by what constitutes a valid block, did not change.

As a consequence, non-standard transactions can still be included in a block because they are valid transactions from a block verification standpoint. However to discourage them, most P2P nodes and most miners ignore them.

NYT has an article claiming Nick is Satoshi.

http://www.nytimes.com/2015/05/17/business/decoding-the-enigma-of-satoshi-nakamoto-and-the-birth-of-bitcoin.html?_r=0

While I agree the evidence shows him to be in the likely list, there is a pool of other people who worked on similar projects and participated in similar forums, any one of which could be Satoshi as well.

Ok. And I see that the mempool (I run a standard bitcoind) is even bigger, but according to docs the nodes keep invalid transactions, and logically they therefore also keep the nonstandard transactions, in the memory pool until shutdown. There seems to be no way to display the number of valid and standard transactions waiting in the mempool in a standard node. Correct?

Edit: After (upgrade and) restart of my node, the mempool is only 107 long, i guess those are mostly valid and standard transactions waiting to get in.

That seems to be correct.

And ... each node holds a different set of non-standard and invalid transactions, because they are only held, not propagated. Correct?

Good question.

I'd define it as Bitcoin is the blockchain data structure and the set of rules that govern what constitutes a valid block. If you want to change the rules that govern what is a valid block (transactions and block structure) then that is a fork. Anything else is not a fork. The reasoning is simple, these are the rules that govern if a specific chain is valid and thus govern consensus.

Nodes however can operate however they want, as long as they agree on the rules that define a valid block. If some nodes decide on a different set of rules, then they will be forced onto their own chain and lose consensus with the main chain.

Take Gavin's IBLT proposal as an example. This is a major change to bitcoind that changes how nodes communicate blocks to each other. This is not a fork simply because it does not change anything at all about what defines a valid or invalid block. Just the communication of blocks.

Take LukeJr's insertion of Ubuntu nodes that block SatoshiDice type services into the Ubuntu default client. This was not a fork because he simply made the nodes decide not to relay transactions to/from specific addresses, but these nodes still followed the same rules on what is a valid block, and so they would accept these transactions after they were confirmed in a block. (To do otherwise would be a fork and they would be kicked onto their own tiny chain).

The stock market seems ripe for a major collapse. Is gold likely to rocket? And bitcoin? Also the dollar has been sagging quite a bit lately. That should help gold but what about BTC?

Compromise solution: up to 500 tps on-chain while keeping the 1MB limit, however, a hard-fork is still needed.

This is an idea which I have been mulling over since Gavin came up with IBLT, but have not expounded because, until recently I did not realise the depth of support within Core dev for maintaining the 1MB. If dev are not able to compromise on a can-kick, e.g. 20MB, or an adaptive limit which leverages the demand for larger blocks with pressure to keep a cleaner UTXO set better fees market, then there is an alternative. Also, Pieter's "headers-first" change which went live in version 0.10 makes this alternative safer on resync.

A "third way": Decoupling the size limit check on blocks from new block messages.

Today, the 1MB size limit is bluntly used to check any block which a node sees. It does not matter whether the block being checked is 2 years old and just read from disk, or whether it is a new block announcement to a fully-synced node. The principle here is that the 1MB check would be retained for new block announcement messages, but not when processing old blocks, or applied after a new block is constructed from an announcement message.

A hard-fork is still required, as the window is then opened for blocks to be accepted which are >1MB on disk, and to support compression/decompression of new blocks.  All nodes will need to be able to decode blocks which might be compressed by different methods.

Notes:

Bandwidth usage for new blocks is kept to 1MB per 10 minutes. People can continue to mine via TOR.

More than one type of compression would then be supported, e.g. IBLT and transaction hash lists, so miners have a choice of which to use. IBLT could be used on top of hash lists which gives even greater compression. CPU overhead is a consideration.

The network can already handle a much higher tps of real-time unconfirmed tx so block propagation methods which rely on the initial publication of tx are encouraged.
e.g. IBLT usage is incentivized for miners wanting to construct blocks larger than 1MB. This also reduces blockchain spam as miners cannot stuff new blocks >1MB with secret transactions. This reduces the rate of increase in UTXO relative to average block size increase.

Fees on a 20MB disk block would approximate the 6.25 reward in 5 years time. This happy state could be reached while retaining the 1MB limit on new block messages.

Resync would still require sending blocks >1MB, however this is not on the critical path for the whole network, just one node.

Current status:

Wladimir has indicated that he wants to close off version 0.11 soon.
If nothing is done about the 1MB in version 0.11 then one of the last chances is wasted for achieving a smooth transition to larger tx volumes without a period of serious confirmation delays, (user unhappiness, bad publicity, price hit, harm to the network effect).

I suggest that if no compromise is possible in dev on the 1MB then at the very least version 0.11 should prepare the way for block compression, and provide an incentive for it to be used, incorporating software to decode an IBLT, and be able to use transaction hashes like the block relay service. This might take a number of weeks/months, but far less time than waiting for Lightning Networks, tree-chains, or other off-chain scaling solutions.

IMHO, the safest hard fork uses both a block version transition PLUS a 30 or 60 day grace period. i.e. after the 75% target is reached, a delay occurs before the change becomes effective. This achieves mining consensus and also allows non-mining nodes and laggard miners more time to upgrade.

Have at it!