[Bitcoin-development] Chain pruning

[Gregory Maxwell]

On Thu, Apr 10, 2014 at 3:24 PM, Jesus Cea <jcea at jcea.es> wrote:
> On 11/04/14 00:15, Mark Friedenbach wrote:
>> Checkpoints will go away, eventually.
> Why?. The points in the forum thread seem pretty sensible.

Because with headers first synchronization the major problems that
they solve— e.g. block flooding DOS attacks, weak chain isolation, and
checking shortcutting can be addressed in other more efficient ways
that don't result in putting trust in third parties.

They also cause really severe confusion about the security model.

Instead you can embed in software knoweldge that the longest chain is
"at least this long" to prevent isolation attacks, which is a lot
simpler and less trusting.  You can also do randomized validation of
the deeply burried old history for performance, instead of constantly
depending on 'trusted parties' to update software or it gets slower
over time, and locally save your own validation fingerprints so if you
need to reinitilize data you can remember what you've check so far by
hash.