James Hilliard via bitcoin-dev
2018-01-01 18:50:04 UTC
Recently a large merchant payment processor has decided to drop
support for BIP21 payment URI's in favor of accepting exclusively
BIP70 payments which has brought to light a number of problems with
BIP70:
1. Many wallets do not support BIP70 and have no near term intention
of doing so.
2. BIP70 requires large complex PKI dependencies such as X.509 and TLS
support(usually via openssl) which have a large attack surface and
poor track record when it comes to vulnerabilities.
3. Signing transactions with keys resident in the same application as
that which handles TLS greatly increases the possibility of keys being
leaked due to vulnerabilities in TLS libraries such as
openssl(heartbleed etc).
4. Sending payments first to a BIP70 compatible wallet before sending
to the merchant increases fees and uses more block space than sending
directly since it is often not feasible for users to fully migrate
funds to a BIP70 compatible wallet.
5. Paying a BIP70 invoice with an incompatible wallet currently
requires manual non-user-friendly workarounds such as
https://github.com/achow101/payment-proto-interface
I propose that we move the BIP70 protocol implementation into a
browser extension that can communicate with wallets over a simple IPC
mechanism such as
https://developer.mozilla.org/en-US/Add-ons/WebExtensions/Native_messaging
in addition to acting as a translation layer that can convert BIP70
URL's into standard BIP21 URI's for wallets that do not wish to
support BIP70 or other custom schemes.
This will provide a number of advantages over the current method of
implementing BIP70 directly within wallets:
1. It removes complex/risky dependencies from wallets and moves them
into the browser which already has to implement full PKI support.
2. It re-enables payment support for wallets that only support
BIP21/normal addresses.
3. It makes offline/custom signing schemes easier to use with BIP70.
support for BIP21 payment URI's in favor of accepting exclusively
BIP70 payments which has brought to light a number of problems with
BIP70:
1. Many wallets do not support BIP70 and have no near term intention
of doing so.
2. BIP70 requires large complex PKI dependencies such as X.509 and TLS
support(usually via openssl) which have a large attack surface and
poor track record when it comes to vulnerabilities.
3. Signing transactions with keys resident in the same application as
that which handles TLS greatly increases the possibility of keys being
leaked due to vulnerabilities in TLS libraries such as
openssl(heartbleed etc).
4. Sending payments first to a BIP70 compatible wallet before sending
to the merchant increases fees and uses more block space than sending
directly since it is often not feasible for users to fully migrate
funds to a BIP70 compatible wallet.
5. Paying a BIP70 invoice with an incompatible wallet currently
requires manual non-user-friendly workarounds such as
https://github.com/achow101/payment-proto-interface
I propose that we move the BIP70 protocol implementation into a
browser extension that can communicate with wallets over a simple IPC
mechanism such as
https://developer.mozilla.org/en-US/Add-ons/WebExtensions/Native_messaging
in addition to acting as a translation layer that can convert BIP70
URL's into standard BIP21 URI's for wallets that do not wish to
support BIP70 or other custom schemes.
This will provide a number of advantages over the current method of
implementing BIP70 directly within wallets:
1. It removes complex/risky dependencies from wallets and moves them
into the browser which already has to implement full PKI support.
2. It re-enables payment support for wallets that only support
BIP21/normal addresses.
3. It makes offline/custom signing schemes easier to use with BIP70.