Re: dpkg binary dbase (was Re: Debian vs. Red Hat)

[loki <eloki@dingoblue.net.au>]

On Wed, Sep 06, 2000 at 07:02:21PM -0500, John Hasler wrote:
> Chris Gray writes:

>> I understand that dpkg is a much easier tool to use.  It is also a
>> lot slower.  It would be nice to write it with a binary database.
> 
> _NNNNNNNNNNNNNNNNNOOOOOOOOOOOOOOOO_
> Ahhm. 
> Do you want to try to edit a binary database to fix screwups?

I thought Solaris used binary databases for speed, with a text one
as backup and for readability.  What if we had both a text and
binary database, and added the following options to dpkg:

dpkg --use-text-avail        Use plain text available
dpkg --use-binary-avail      Use binary available
dpkg --gen-binary-avail      Generate binary db from the text version

At least this way, the choice is up to the user (with a suitable or
configurable default).  The binary database can be used for speed,
but we can always regenerate it from a given text database (which is
fixable) if it gets screwed up.

Pros: would, presuambly, be faster
Cons: extra complexity, and extra disk space used to store 2 forms
of the same data (but not everyone may generate a binary db if it's
optional).

I guess the question is how much speed we would gain, and whether
it's worth the cost in complexity and space.



-- 
    loki
eloki@dingoblue.net.au

Dare I disturb the universe?  You bet I do! :)