Subject: RE: Database of DID's (start of a solution?)
From: Jonathan Newman (jnewman@mudpup.com)
Date: Thu Dec 07 2000 - 23:59:03 EST
> How about a database for JUST the major/minor parts of the the
> identifier. If this is limited to 256 entries (8 bits of index), it
> would allow quite a bit of mounted volumes. Then there would be no
> "evil hash". OK, so it takes a table in memory that has
> the accumulated
> major/minor numbers, and it needs to be written out, but after it is
> started, it should be pretty much self sufficient. Since the DID is
> a total of 4 bytes, this allows 16megs of unique inodes for
> a volume.
> In my book, this sounds pretty adequate.
The only reason for using the dev/inode pair was that it is a
semi-permanent way to enumerate the directories. A database wouldn't
have to depend on those, and would certainly be more portable if it
didn't -- if you move your contents to a new drive/volume, your
database shouldn't become invalid. I think something more complicated
that just the device number is needed. The lastdid solution would
work quite well if it only persisted between runs of afpd.
Jon
This archive was generated by hypermail 2b28 : Wed Jan 17 2001 - 14:32:45 EST