On Wednesday 07 January 2004 04:45 pm, Dan Goodes wrote: > One of the big issues of course is that a [large] number of people will > install the kernel-source RPM so they can make changes/patches, e.g. to > get the latest XYZ driver that isn't bundled with the kernel yet. If I > understand the above, it would become a painstaking process to re-patch > and re-compile the kernel completely from scratch each time we install an > update via this "delta" mechanism. And how is this different than what we have now when the kernel-source RPM is updated via up2date? The rhmask process, again IIRC, took the pristine RPM for whatever package, downloaded the errate delta, applied the delta to the upacked pristine RPM, and produced the full errata RPM as its output. You then installed the errata RPM as if you had downloaded it whole. Specifically: 1.) You download kernel-source-2.4.22-1.2138.nptl.rpm.fc1.rhmask 2.) You get your pristine copy of kernel-source-2.4.22-1.2115.nptl.i386.rpm 3.) You run rhmask kernel-source-2.4.22-1.2115.nptl.i386.rpm kernel-source-2.4.22-1.2138.nptl.rpm.fc1.rhmask 4.) Install the resulting kernel-source-2.4.22-1.2138.nptl.i386.rpm just as if you downloaded the whole thing. But I said something _LIKE_ rhmask. The maskfiles weren't optimized for transmission efficiency; a similar mechanism could be used. Rsync can already do this, you just have to keep the original RPM's lying around. Xdelta would work just as well. And I really should not have used the term delta for what rhmask is using, since it is an XOR maskfile, which is a very primitive form of delta. A better delta would not be between the complete RPM packages, but a package of the individual file deltas. -- Lamar Owen Director of Information Technology Pisgah Astronomical Research Institute 1 PARI Drive Rosman, NC 28772 (828)862-5554 www.pari.edu
