[cl-store-devel] optimizing default backend

[Max Mikhanosha]

Is there any interest in having following type enhancement to default
backend, and will a patch implementing it be considered?

* Introduce buffering by an having an (UNSIGNED-BYTE 8) array
  bound to a global (similar to the circularity hashes)

* Replace every WRITE-BYTE with an inline wrapper that buffers into
  the above array, and calls a generic FLUSH-STORE-BUFFER when its full

* Default FLUSH-STORE-BUFFER method will call write-sequence

Above was brought on by disappointing timings of using CL-STORE to
serialize to files vs serializing to in-memory streams under
SBCL. Tried both flexi-streams and babel-streams, and it seems writing
to files directly is around 3x times faster then writing to in-memory
streams

Regards,
  Max