This month's State of Bitcoin Address issued by the Bitcoin Foundation largely covers the submission and testing of two patches which promise to greatly reduce the memory usage of the Bitcoin reference client. The two major patches submitted are named "Orphanage Thermonuke" and "Transaction Orphanage Amputation" and in tandem they have the potential to seriously clamp down on Bitcoind's memory footprint and could lead to Bitcoin nodes running on more kinds of hardware. The patches work in tandem to drastically alter the Bitcoind client sync mechanism in a radical way.

Its author asciilifeform describes the effect of the Orphanage Thermonuke thusly:

The patch below shoots the entire 'orphan' (bastard) block storage mechanism in the head. To force sync, we simply recurse every time such a block is encountered, asking for a parcel of blocks preceding it, down to the current best-known.

Bastard blocks themselves are discarded.

Testing on Bitcoin versions 0.5.3 and the foundation's newer 0.5.3.1 has consistently revealed that during initial sync a major reason for expanding memory consumption and Out Of Memory kills was the collection of unverifiable blocks whose predecessors in the chain had not been downloaded leaving them to collect in RAM. Foundation co-chair mod6 in the report mentioned that doing a full sync test of Bitcoin v0.5.3.1-RELEASE with the Orphanage Thermonuke patch applied he avoided any OOM kills and has detailed test results on the full synchronization process.

The second patch "Transaction Orphanage Amputation" depends on the Thermonuke patch, but goes a step further to prevent the accumulation of any orphan transactions in RAM which can not be verified against a previous transaction.

Testing of the new patched Bitcoind variations against their predecessors continues apace as the Foundation's project of tearing the Bitcoin reference implementation down to the bare wood appears to be reaching the point where their software is concretely improved over its predecessors. Given the amount of further testing the reports indicates as necessary, the Foundation has not offered a target for when they might offer a version 0.5.3.2 release.