1. 14
  1.  

  2. 4

    A year or so back I met an engineer who proudly proclaimed his principle accomplishment that he got the Midori project killed. I think he expected awe or praise as opposed to the cocked eyebrow and mild look of disbelief he got from me.

    1. 3

      Did he explain the reasoning for having done so?

      1. 3

        …what was his reasoning?

        1. 4

          From what I briefly gathered, a conviction that operating system level stuff had to be done in unmanaged code. The fact that Midori demonstrated otherwise apparently didn’t matter.

          1. 3

            It didn’t really. One of the problems with Midori was that it had more unmanaged code providing the CLR than most microkernels had in total. The size of the unmanaged TCB was very large. In other respects it suffered primarily from bad timing. Turning Windows into a libos added about 30 MiB of RAM overhead per app. That was a lot at the time, now it’s noise (and probably would have been largely offset by the savings from the large amount of safe sharing possible between system services).

            1. 1

              I’m not sure that’s proof that it doesn’t work. Those are costs, but they seem similar to the arguments against microkernels because of performance.

      2. 1

        Awhile back I was learning about Singularity OS on which Midori was supposedly based. Excited to read this.

        Stories with similar links:

        1. Blogging about Midori via adsouza 7 years ago | 18 points | no comments