How many TXNs in the mempool could be bundled.

[d4n13]

I know the subj is confusing... read on...  I gather fees are a basic supply/demand equation.  Supply is the space in the block.  Supply can be increased by using Segwit and bundling transactions.  I'll leave LN for now because as of today (Jan2018) there are very few wallets that support it.  Now, I may be able to bundle some of my TXNs buy leaving my coins on my phone longer than I want so that when I move them to cold storage I do it monthly instead of weekly.  But I think the biggest opportunity for bundling would be the exchanges.  Specifically on outbound TXNs.  All the exchange TXNs I've done are the typical <5 input and <5 output TXNs.  I'm certain this could be optimized, but since some of those fees are paid by the user, the exchanges incentives are one degree removed, slowing adoption.

So.. If my suspicion is correct that a significant number of TXNs are unbundled exchange TXNs, has anyone put some serious study behind it.  Obviously we are talking millions in fees here, so that is a lot of money to leave on the table.  It would seem that if you knew an exchanges hot-wallet, thier mining pool and some possible other tells, it would be possible to build a model where there is some statistical confidance that TXN y belongs to exchange x.  For the rEDITORs out there, I did say statistical as in less than 100% and more than 0%.  Not certainty.

If I'm still on the tracks, I'd be curious what fees would look like in this utopian world (even without LN) where we do effectively the same volume, but all exchanges use segwit (from addr) and their TXNs are bundled.  Bundles could be every hour (less optimized), or every day (most optimized).

TLDR: Fees would be negligible if exchanges bundled and used segwit.  It would drop fees by {TBD}%

[Youghoor]

f I'm still on the tracks, I'd be curious what fees would look like in this utopian world (even without LN) where we do effectively the same volume, but all exchanges use segwit

And you are wrong, only 1 of 5 exchanges are using segwit addresses.

Well it touched more than 250k a few days ago when there was another attack from Bcash in the main blockchain.

I would not be surprised if the whole mempool receives more than 300k or 400k in there, and no, it is not going to collapse because of that.

We all know that bitcoin has been created in order to handle max 20 transactions per second, and now it is handing more than 98 per second, at the same time.

[hatshepsut93]

You are right and I think some big exchanges like Coinbase are not doing it on purpose - they want Bitcoin to suffer so altcoins that claim to be scalable can enjoy some pumps - for exchange this would mean more fees for trading them, they also might be bribed to do so by those altcoins, and exchanges would hugely benefit from network centralization that comes with bigger blocks - they would be able to run nodes while most home users will be kicked out of the network, and with being one of the few nodes of the network comes the ability to dictate the rules when joined with miners as well as censoring transactions of SPV clients. So, it makes a lot of sense that some exchanges really wanted SegWit2x to succeed. And what is worse, this attacks doesn't really cost them much, because people are not leaving since there are very few alternatives.

[d4n13]

And you are wrong, only 1 of 5 exchanges are using segwit addresses.

Actually wasn't aware that any exchange was using segwit.

Did some envelope math and I think bundling would save between 2-4% in TXN size/cost.  Not much, but some.  Segwit obviously saves much more than that.