Should we remove tx_extra?

xmrenthusiast

I have seen the feature discussed in multiple places (Bitcointalk, Reddit, Stack Exchange, Dev meetings, etc) over several years.

My understanding is that:

There are some potentially negative privacy implications of using tx_extra because of possible fingerprinting from a rarely used feature.
The use of tx_extra is still relatively uncommon after years of availability to a growing base of Monero users and developers.
Most theoretical use cases such as refund transactions or messaging can be solved with other communication tools. In the case of Zcash use of the encrypted memo field with applications such as Zbay, appear to be result from a strong secondary motivation (to increase shielded transaction usage from a tiny base). Thankfully this is not as relevant to Monero where privacy is present by default. In fact the fingerprinting that results from tx_extra on balance appears harmful to network privacy.

I would love to here more convincing arguments about why the tx_extra field should be retained. Several of the better arguments I have seen discussed on GitHub relate to potential soft fork utilization and (theoretical) future development availability here and here. This padding idea appeals to me as a compromise (privacy improves while the feature remains) but must be balanced against the cost of blockchain bloat.

sgp_

This is a really good conversation to have. I think in any case no matter what decision is chosen, users need to be informed of the proposed change in advance and have plenty of time to leave arguments for or against. Even though this feature isn't currently being used by many, it's the removal of a large set of potential things people can do with the network.

jtgrassie

Bear in mind, tx_extra is used by (at least):

every tx for the tx shared secret
pools and miners as extra nonce
certain other txs (like multisig and multi-dest subaddress txs)

Thus "removal" is not possible without catering for alternative ways to accomplish above. That's not to say we shouldn't discuss, just that it would be a big change affecting both the Monero codebase and other things like pool/miner implementations.

There have been a couple of alternative suggestions to blanket removal of tx_extra, such as a fixed (or quantized) size with encrypted data, though that of course has other implications (bigger txs, slower parsing etc).

Low-hanging fruit for removal of another possible fingerprinting "feature" would be removal of the unlock_time.