server: increase connection stack size from 256K to 512K for vtwrite
server: vtwrite() of existing data
server: detect if written data already is present on the server
server: handle partial packets, write call shim
mk: disable profiling by default
disk: handle map parsing failure correctly
cache: use atomic counter for initial buffer allocation
cache: keep entries locking during reads
cache: random eviction when full
cache: load factor is definitionally always < 1
fix two warnings; thanks wilhelm ☺
cache: restore bucket locking; fixes multiclient reading
disk: don't free stack buffer on decompression failure >_<
server: give each client its own Proc
server: fix connection cleanup
server: revert multithreading fix. buggy.
server: handle each client in a separate proc
disk: return error to client on read failure
index: report bucket lookup error to client
fuck it, drop all of the first pass
old: drop the first pass at config parsing
cache: loop around buckets when probing
cache optimizations; 2x faster than venti!
disk: remove needless allocation/free in the index access path
handle each client in its own proc
report errors as coming from neoventi
neoventi: initial commit. Working read-only!