1. 17
  1.  

  2. 1

    Every time you see an open-source rewrite of a system from a Google paper, you should keep in mind that Google wouldn’t have published the system unless it had already been replaced by something better.

    1. 16

      This is not true at all.

      1. 6

        Spanner, Zanzibar and Borg among outers are proof that this isn’t true. GRPC is even a little different that they are still adopting it as it was released publicly.

        Systems are going to change over time, so in the long term there will be outdated papers, but that is just a fact that papers are a snapshot of knowledge at the time of authorship.

        1. 4

          Even if they have, so what? If the concept is one that (when implemented) works well, and there isn’t an obviously better alternative publicly available, who cares whether or not it’s not pulled straight from Google’s latest stack?

          1. 4

            Though, even if Google’s moved on internally, these implementations can have a big impact - Hadoop comes to mind.