[bitcoin-dev] A new payment address format for segregated witness or not?

[jl2012 <jl2012@xbt.hk>]

On the -dev IRC I asked the same question and people seem don't like it. 
I would like to further elaborate this topic and would like to consult 
merchants, exchanges, wallet devs, and users for their preference

Background:

People will be able to use segregated witness in 2 forms. They either 
put the witness program directly as the scriptPubKey, or hide the 
witness program in a P2SH address. They are referred as "native SW" and 
"SW in P2SH" respectively

Examples could be found in the draft BIP: 
https://github.com/jl2012/bips/blob/segwit/bip-segwit.mediawiki

As a tx malleability fix, native SW and SW in P2SH are equally good.

The SW in P2SH is better in terms of:
1. It allows payment from any Bitcoin reference client since version 
0.6.0.
2. Slightly better privacy by obscuration since people won't know 
whether it is a traditional P2SH or a SW tx before it is spent. I don't 
consider this is important since the type of tx will be revealed 
eventually, and is irrelevant when native SW is more popular

The SW in P2SH is worse in terms of:
1. It requires an additional push in scriptSig, which is not prunable in 
transmission, and is counted as part of the core block size
2. It requires an additional HASH160 operation than native SW
3. It provides 160bit security, while native SW provides 256bit
4. Since it is less efficient, the tx fee is likely to be higher than 
native SW (but still lower than non-SW tx)
---------------------------

The question: should we have a new payment address format for native SW?

The native SW address in my mind is basically same as existing P2PKH and 
P2SH addresses:

BASE58(address_version|witness_program|checksum) , where checksum is the 
first 4 bytes of dSHA256(address_version|witness_program)

Why not a better checksum algorithm? Reusing the existing algorithm make 
the implementation much easier and safe.

Pros for native SW address:
1. Many people and services are still using BASE58 address
2. Promote the use of native SW which allows lower fee, instead of the 
less efficient SW in P2SH
3. Not all wallets and services support payment protocol (BIP70)
4. Easy for wallets to implement
5. Even if a wallet wants to only implement SW in P2SH, they need a new 
wallet format anyway. So there is not much exta cost to introduce a new 
address format.
6. Since SW is very flexible, this is very likely to be the last address 
format to define.

Cons for native SW address:
1. Addresses are bad and should not be used anymore (some arguments 
could be found in BIP13)
2. Payment protocol is better
3. With SW in P2SH, it is not necessary to have a new address format
4. Depends on the length of the witness program, the address length 
could be a double of the existing address
5. Old wallets won't be able to pay to a new address (but no money could 
be lost this way)

------------------------------

So I'd like to suggest 2 proposals:

Proposal 1:

To define a native SW address format, while people can still use payment 
protocol or SW in P2SH if the want

Proposal 2:

No new address format is defined. If people want to pay as lowest fee as 
possible, they must use payment protocol. Otherwise, they may use SW in 
P2SH

Since this topic is more relevant to user experience, in addition to 
core devs, I would also like to consult merchants, exchanges, wallet 
devs, and users for their preferences.