>At the moment (for alpha at least, and probably similiar for other
>architectures unless something really funny is going on) there would
>appear to be in the order of about 400 different system package names
>making up a full distribution.

um...wha?

>On my PIII 700MHz laptop, 'pkg_info | wc' returns 155 and takes an
>non-insignificant amount of time (fivish seconds - didn't time it) the
>first time it's run (of course it's nearly instantaneous the second
>time).  Think of that on an old vax or m68k box.  And the database for
>those 155 packages use just over 8MB of disk and 1340 inodes, so we're
>talking perhaps over 3000 inodes and perhaps 20MB used to keep system
>package info.  This seems a bit excessive...

my sparsely used sparc has 17 pkgs.  that's one open() on /var/db/pkg,
17 chdir() calls, and 17 more calls to open() (for the +COMMENT
files).  this might be a place where the ffs practice of allocating
inodes for directories not so close to each other tends to bite.

>Note - these figures are a very rough.  I haven't looked into things
>like my average installed package size (in terms of number of files)
>versus the average system package and so on.

i've also noted that it's slow the first time and faster after that.
somehow i rationalized it away...

certainly lintpkgsrc seems like it ought to be faster.  :)

-- 
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org             * "ah!  i see you have the internet
twofsonet@graffiti.com (Andrew Brown)                that goes *ping*!"
andrew@crossbar.com       * "information is power -- share the wealth."