b+tree + bloom status

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Robert Collins wrote:
| Cool.
|
| So, groupcompress does not read a compression graph to get keys out.
|
| So its closer to torture_miss than to get_components. I think we'll want
| to profile blooms with groupcompress separately to knits.
|
| -Rob

Sounds like we would. It might actually do *better* there, considering
you are requesting a single key, rather than a bunch of keys. The
problems with a bunch of random keys was always that we needed to read
all the pages anyway.

I was also thinking that miss torture might suffer from that as well.
When you are looking at 100,000 keys => 1,000 key node, even though
you've filtered 99%, there are still 1,000 keys that are false positives.

Miss torture might be a bit *too* much versus a real-world workload.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkh/ZyAACgkQJdeBCYSNAAP22gCfVk3rRn70F3lUP3hhaDhdapIB
DwEAoMRZcQYOEkysAMgEyYzSEfCm4cUz
=nuP5
-----END PGP SIGNATURE-----