[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: debian-trivia

On Mon, Jan 06, 2003 at 06:02:58AM -0800, Alexander Hvostov wrote:
> Perhaps it's high time dpkg received some serious optimization. The way
> in which the package control info is stored and processed is horribly
> inefficient.

Yes.  (I run Woody on a 486DX2/66, 16MB RAM.  dpkg crawls.)

> Here's some suggestions:
> - Rather than storing data for each and every package in the single file
> '/var/lib/dpkg/{status,available}', store each package's data in its own
> file '/var/lib/dpkg/{status,available}/foo'.

IMHO that would be *less* efficient.  Ever noticed how long it takes to
ls /var/lib/dpkg/info?  Ext2/3 does not cope well with lots of small
files in a big directory.

> - Store frequently accessed control fields from status/available in a
> cache (Berkeley db or some such) for fast access. The cache is, of
> course, rebuilt if deleted or /var/lib/dpkg/{status,available}/* changes
> (as happens with 'dselect update').

I think that is a good idea.

> - Store a database of files on the filesystem and the packages they
> belong to in a cache, like above.

I believe dlocate does something like this.

Marius Gedminas
CBQ is merely the oldest kid on the block - yet it is by far the least useful
qdisc and also the most complex one. I advise *against* using it. This may come
as something of a shock to many who fell for the 'sendmail effect', which
learns us that any complex technology which doesn't come with documentation
must be the best available.
		-- Linux Advanced Routing HOWTO

Attachment: pgplPmWERKrHQ.pgp
Description: PGP signature

Reply to: