Re: apt (was: Re: [Plug] redhat 6.1 insllation)

["Michael C. Toren" <mct@netaxs.com>]

> >  apt-get install package1 package2 package3
> > 
> > And it'll install the latest version of each package you specify.
> 
> I don't think this motif would work under the current RPM specification.
> By definition, the filename of a package has *no* significance.  The
> package name, version, and release are all defined within the file (in the
> md5sum'ed part of it, I might add).  Does apt-get rely on a database
> relating file names to actual packages?

Prior to doing an "apt-get install N", you need to do an "apt-get update",
which will read through your list of sources (which is simply a list of
http://, ftp://, or file:// URL's) and build a current list of the packages
(and versions) available.  When you then say "apt-get install N", it looks
through the cache to determine which files are needed, and fetches them.
What's even cooler is that each package listed in the cache also has a
list of packages it depends upon, suggests, conflicts with, replaces, etc.

Because the cache of available packages also includes version numbers,
it's possible for apt-get to build a list of packages available with newer
versions than you currently have installed.  "apt-get update" generates
such a list, and fetches the newer versions.

It's a very, very nice system.

(Also, the name given to a package is defined in its control file.  So,
yes, it's possible for a file foo.deb to contain a package named bar)

-mct