On Sat, Oct 8, 2016 at 2:34 PM, Julian Andres Klode <j...@debian.org> wrote:
> On Sat, Oct 08, 2016 at 02:15:27PM +0200, Michael Stapelberg wrote: > > Package: apt > > Version: 1.3 > > Severity: normal > > > > Recently (sorry, I don’t know with which version precisely), I have > > noticed that apt acquires pdiffs when running “sudo apt-get update” > > (yes, I have updated my aliases to prefer apt over apt-get for the > > future). > > > > Since pdiffs make updates way slower in my setup, I have pdiffs disabled > > on all my machines using: > > Are you running everything from a super hyper fast on-site mirror? > Otherwise > I can't believe that statement. PDiffs are tiny, and apply at an incredible > speed, in parallel. And with 1.3, PDiffs are mostly applied while other > indexes are downloading (if any). > I’m on a 1 Gbps fiber connection and can max out that connection to the closest mirrors (e.g. ftp.ch.debian.org, or deb.debian.org via fastly). In this setup, even uncompressing files is usually a slow-down over just downloading the uncompressed version. Applying diffs is usually even slower. That said, I was under the impression that my observations hold true for slower connections as well, so I’m intrigued. > > In any case make sure to time things. Depending on the repository, pdiff > in 1.3 is much faster than before, and the 1.2 pdiff is already much faster > than the 1.1... > What’s a good way to measure this? If there’s no canonical way, I’ll restore /var/lib/apt/lists from yesterday’s backup and use that for benchmarking. > > Apart from that, see what Niels said. Site-local settings are best > stored in files starting with 99 IMO. Thanks, I’ve changed all my files to start with 99. -- Best regards, Michael
