1. 15
  1.  

  2. 8

    I remembered seeing a response to this, and dug it up: http://harmful.cat-v.org/software/andy_tanenbaum

    1. 2

      It was a few years after this that L4 was developed, and it wasn’t too long after that CYGWIN was initially released. It’s funny to see how quickly some of these things change.

    2. 7

      How would one use these questions to evaluate the trend of microservices running in docker?

      I’d say we have shifted everything a level up (or down?) but are basically rebuilding the same things as 20 years ago.

      1. 5

        I would say it’s given some credence to “UNIX can be successfully run as an application program”, at least.

        Where I think the modern stack differs from 1992 is:

        Distributed shared memory in one form or another is a convenient model

        I think this is largely untrue, unless you consider a database an example of distributed shared memory. But the Gemstone) or Terracotta approach of trying to make 100 machines think they all share the same RAM is mostly dead, IME.

        I’d say that a lot of dist systems being written right now are basically a victim of 80% being good enough. I think a system being distributed is not really appreciated by many authors. For example, I still see people asking for the ability to nicely flush requests before restarting a process handling requests rather than accepting that a machine can die at any point and pushing retry logic into clients. There are also very popular tools like Cassandra that use timestamps for global ordering and throw away the concept of causality in a distributed system.

        My two cents is that a lot (not all) of the docker fetishism has more to do with valuations being tied to the size of a company rather than other metrics. So companies have to become big and trying to figure out how to be productive. The pessimist in me says that most of the “advances” in dist sys that has happened over the last few years is mostly driven by fashion or organization limitations rather than pushing the state-of-the-art.

        1. 2

          Yeah, shared ram is out. Replaced by distributed “cap defying” databases.

      2. 6

        This post is from 1992. Adding a date might stop some knee-jerk reactions.