On Fri, Aug 27, 2004 at 07:50:34PM +0200, Michelle Konzack wrote: 
> Too blown...

XML of course, but solutions like SQLite could be worth investigating.

More generally, which data structure/access method does apt/dpkg use to
query the Packages files? (yes, I'm ready to read replies like "go and
read the code" . I guess the whole Packages database is read in memory
at startup, stocked in some data structure which is accessed when
needed.

In this way the work of creating the data structure has to be performed
each time apt/dpkg are run. Using a database (whatever: sqlite, mysql,
dbm files, ... I'm just discussing the architecture) the data structure
would be built on update actions and needs not to be rebuild each time
apt/dpkg are run, it would just be queried.

Of course in this way I'm not adressing the bandwith issue, but only the
processing time.

Regarding bandwith I think the right way to go is to use diffs, our
Packages file are relatively stable and I guess that between upgrades
only a minor part of them changes. We can both consider distributing
incremental packages, but I think it would be easier to just perform
diffing on the network and make apt downloads just diffs using for
example HTTP Range headers (or whatever, I don't remember the exact
name).

Cheers.

-- 
Stefano Zacchiroli -*- Computer Science PhD student @ Uny Bologna, Italy
zack@{cs.unibo.it,debian.org,bononia.it} -%- [url]http://www.bononia.it/zack/[/url
]
If there's any real truth it's that the entire multidimensional infinity
of the Universe is almost certainly being run by a bunch of maniacs. -!-

[ Post a follow-up to this message ]