Vitalik just got finished describing SegWit as basically an ugly kludge of code. Doesn't he know about the transaction malleability? Mt. Gox? Kittens??
There is no need for that ugly kludge -- split transactions and blocks into two records -- to fix transaction malleability. It would sufflice to skip the signatures when computing the transaction id.
Blockstream's reasons to want that kludge, ignoring all objections, are obscure. It is not necessary or helpful for fixing malleability, and does not reduce bandwidth or storage costs. On the contrary, there are alternative solutions to save bandwith from miners to clients that are simpler and more effective.
One possiblility is that they want the freedom to muck around with the signatures withot having to justify or explain to anyone, since they could claim that the "main" record contains the information that other ordinary wallets need, while the contents of the extension block neeed to be understood only by them.
Ot perhaps the LN will require some horrendoulsy complicated signatures; then SegWit would be a way to accomodate such transactions without impacting the bandwidth or requiring an an increase in the block size limit. Ans maybe also a way to keep the LN fees down: Pieter suggested that the fee rate (mBTC/kB) for the signature record would be a fraction of that of the main records, ostensibly to encourage use of the SegWit format.
So assuming there are no scaling advantages, what is the point exactly of separating signatures from transaction chains or whatever?
There is a modest scale advantage, but it is equivalent to a one-time increase in the block size (different numbers have been cited as for how much, depending on various assumptions). There is no scaling advantage, as it doesn't help with ongoing increases in bandwidth, storage, or anything else.
The benefits are:
1. Can introduce a new signature format.
2. Fixes malleability
3. New nodes don't need to download signatures below a checkpoint, since they aren't going to verify them anyway.
4. One time effective block-size increase.
All of these could also be achieved other ways.