Built-in soft deletes are one of the reasons I like XTDB
---
RT @xtdb_com
https://news.ycombinator.com/item?id=30991628
"life when you don't have bitemporality" - @malcolmsparks
#redditdown
https://twitter.com/xtdb_com/status/1513618269315473416
@icedquinn Yep. A beautiful artifact of immutability. I have never seen them market that fact, but as someone who has gone to pains to implement soft-deletes in half a dozen PostGres systems after wishing we had it, this default with XTDB is nice.
@icedquinn Yeah, I totally get that. But it seems like space concerns are going we way of hierarchical directory structures; that is, kids these days don't know anything about them.
The reason this is generally not done is because the security of any one chain is only as large as the number of participants on that chain (depending on the type of validation done that may be PoW, PoS or whatever). Therefore the challenge is to ensure you dont partition your chains to the point that any one chain can be broken through a 51% attack. Regardless any partitioning at all of this nature will always be a trade off between some degree of security and efficiency (of which space efficiency is pat of that).
Its not changes to the history im worried about. In a 51% attack you can effectively issue brand **new** transactionns on the side chain that can get validated when they shouldnt be. Though again there are solution.
Yes light nodes contribute **nothing** to the security of the network. Their main purpose is just so you can have a local wallet going without needing 100 TB hard drive :)
Yea but conning them into bieving a fake history doesnt actually constitute an attack since they arent the ones validating the history. You might be able to convince them they have a balance they dont have if you overtake a majority of their peers, but since they arent validating transactions themselves that wont do anything useful since they will discover it was all a lie the second they try to act on it (Such as accepting or sending money that doesnt really exist).
if you tried to shove new entries in the 1,000 log, anyone could still check that the master log actually hashes to that.