>Well, yes, that's definitely not optimal. However, there are a couple of
>things
>to bear in mind: (1) the actual cost of this insanity is relatively
>small (which
>certainly surprised me), and (2) doing it this way allows certain
>optimizations
>elsewhere in the packaging system, so that doing it some other way has a
>cost
>that needs to be quantified. Handling a thousand separate files (as an
>example)
>involves more overhead than looking after just the one and - as
>currently
>structured - installing a package involves checking all the contents of
>all
>currently installed packages. Currently, that would be significantly
>more
>expensive than the current lunacy, although that doesn't have to be the
>case.


A test I ran showed that a "cold" lookup in the contents file
costs 100x less than the "cold" lookup in 1000 separate files; and that
is really not surprising.

As an estimate of the cost, run "pkginfo" twice.  You'll find it takes
20-30 seconds the first time and only one the second.

Conversely, "wc contents" takes about 1 second the first and second
time.

Since I feel package installation is a rare event, we really need to
optimize for the standard operations and not the install/upgrade.
The latter can be optimized in different ways.

Casper

Reply via email to