Valid outputs need to originate from valid coinbase transaction in valid block. Non-upgraded nodes will never confirm blocks containing tx that spend non-existent outputs ..... for the same reason they won't confirm blocks containing a 21million coin spend. Those transactions--and if mined, those blocks--are simply invalid, ignored, by old nodes. Miners don't even matter in this context. If they keep building on such a block, their chain will be forked off the network.
Unless, of course, node operators en masse uninstalled their node software and reinstalled software that recognized these outputs as valid. i.e. a hard fork.....
It does not matter what non-upgraded nodes do when majority of the blocks are 21block format
Miners don't matter in a hard fork. Some miners could collude to inflate the money supply, but that is no guarantee that anyone will care to transact on their network. They can just be ignored.
Sigh...
You're confusing transaction type (standard vs. non-standard) with validity. Using non-standard transactions can't hide the outputs from old nodes. They won't be "empty blocks"; they will be "blocks full of invalid transactions."
How would old nodes be forced off the network? You're very confused. Since no consensus rules would be broken--only restricted--the 5% are never forced off the network. Their problem is that they may include transactions in their blocks that are invalid under the new soft fork rules, and thus rejected by the network.
It's pretty clear that you don't understand a damn thing about bitcoin.
This Dr. made some metaphor to explain segwit, but his example how transaction works is obviously wrong. So even a Dr. who is giving lecture about segwit does not understand how bitcoin transaction works (or intentionally give misleading information?), how could the rest of the people have any idea what it is?
I'm not clicking that link. I don't know who this "Dr." is or what relevance he has. But it seems like the burden is on you to explain exactly how he is wrong, rather than just stating it as truth.
People like you pretend to understand how bitcoin works, but fall flat on their face. And many more people than that don't even pretend to understand bitcoin. So your premise is stupid to begin with. How about developers? Well:
Blocktrail CTO and BitcoinJS Co-Maintainer Ruben De Vries: Segregated Witness Not Very Complicated
Lawrence Nahum: Bitcoin Wallet GreenAddress Already Integrating Segregated Witness
Breadwallet CEO Aaron Voisine: SegWit Soft Fork First, Block Size Hard Fork Later
Mycelium's Leo Wandersleb: Segregated Witness a Technical Necessity
Electrum Developer Thomas Voegtlin: Segregated Witness Highly Supportable
Core Developer, libbtc Library Maintainer and digitalbitbox Developer Jonas Schnelli: Segregated Witness is Less Than a Week of Work to Test and Deploy
Ermagherd!!!! Segwit is just too complex!!!!