debian install abandoned

Michael Gratton mike at
Wed Dec 11 13:41:55 CST 2002

stephen white wrote:
> On Tuesday, December 10, 2002, at 04:33 PM, Michael Gratton wrote:
>> Funny, I get fully working workstations after doing Debian installs. 
>> Maybe you weren't doing it right?
> Yes I was.

So how do you explain that I get working workstations after doing a 
Debian install, and you don't?

> If you can't install a package and have all the other packages recognise 
> it's there and start using it, there's no integration.

You mean install a third party application that doesn't put its files in 
the right place, doesn't set itself to start up and shut down properly, 
and generally does not play nicely with your system?

How can you honestly expect software developers to cater for every 
single POSIX-like system, to make their software integrates correctly 
with those hundreds if not thousands of variants? They can't. It's up to 
the distro to make the application play nicely, and Debian's packages do 
a damm good job of it.

If you want integration, if you want all the hard work done for you, 
install the software using the distro's package. If you don't want that, 
install it yourself (to /usr/local) and deal with hassles yourself.

> The latest version of Apache from CVS may not be (and isn't) tweaked for 
> Debian packages.

Why on earth are you installing Apache from CVS? It's most likely a 
*very* bad idea to do so. The same will apply to 99% of the packages you 
install on a server, and maybe 90% of the packages you install on a 

I have 859 debian packages installed on my workstation here at work, and 
less than 68 applications I've installed myself in /usr/local. That's 
92%. On one of our staging servers we have 226 debian packages installed 
and 3 applications installed into /usr/local: 99%.

> ls -l /usr/sbin/apache... error = not installed, no error = installed.

How does your computer know to check /usr/sbin/apache to see if Apache 
is installed without keeping a list of files in the package? How does it 
know what version it is? How does it know if a file has changed, and 
hence not overwrite your existing configuration when upgrading, if it 
doesn't keep a checksum for each file?

>> If I wanted to get a list of installed packages by scanning that lot, 
>> I'd be waiting for years. *Because* dpkg keeps it's own database of 
>> installed software, I can get that list within a useful period of > time.
> Bet the filesystem is better optimised for access than the internal 
> database that dpkg is using. If I'm wrong, why not write dpkgfs and 
> become famous?

That's why Oracle uses the filesystem to store it's data and not some 
custom, stupendously fast storage mechanism on a raw partition. Oh, 
wait, it doesn't.

> You've never had to create a dummy package? Have you only ever used 
> Debian's packages then?

Never created a dummy package. Ever. And as I said, I have 68 non-Debian 
software packages installed. I *do* have reasonable expectations, 
however, of how these software packages will be integrated with my 

> Not really. If Gentoo can compile Gnome source, there's no difficulty 
> for the end user there.

Yes, but it compiles and installs and integrates correctly *only* 
because someone who contributes to Gentoo put the work in to make sure 
that it does. The same goes for Debian packages. The only difference is 
that Debian does the compiling for you, which saves time and usually 

> It might help to understand that I'm pointing out a common deficiency, 
> not championing any one distribution over another. I think all the 
> distributions suck, all Unixs suck, and all computers suck.

Look, I agree with you here, but given this is what we have to work 
with, you really need to either adjust your expectations or just stop 
using computers.


Mike Gratton <jabber:mjg at> <>

LinuxSA WWW: IRC: #linuxsa on
To unsubscribe from the LinuxSA list:
  mail linuxsa-request at with "unsubscribe" as the subject

More information about the linuxsa mailing list