Update patches versus full update RPM

[alan]

On Fri, 9 Jan 2004, Steven Usdansky wrote:

> As an end-user, I like the concept of update patches. I'm willing to keep a set
> of reference rpms handy, or pay the penalty and endure the download of such if I
> don't. 

I don't know if patches will work very well.

In order for the patches to work, you will need to have diffs of every 
version out there.  The update mechanism will have to have logic to be 
able to tell what version you have and what needs to be patched.  This 
becomes more interesting when part of the update has a script to fix 
another problem caused by a previous patch.  

rhmask will not work since the mask is as big as the rpm and requires the 
original rpm.  (rhmask was created to distribute patches for commercial 
apps where you only want someone with the original rpm to be able to use 
the patch.)

There are also other difficulties to take into account...

You have to determine if the original binaries have been modified.  
(Recompiled from source with a different version of the compiler and/or 
different libraries may generate different binaries of the same program.) 
Matching MD5 or SHA1 checksums would give you this information.  

It is possible to do, but it would require a lot of verification and code.  
It is not as simple as it might appear at first glance.