sockets broke that. ioctl() always felt like special magic to me, but
sockets+ioctl is not just special magic, its arcane.
I think part of the problem is that a network binding actually isn't
"just like a file" because it incudes things like out of order
delivery, and control (side) channel.
TBF, sparse files and network file systems is another place things
broke down. For some reason, device-special and sparse files wound up
being in the too hard basket for network mounts.
I know Lindsay Marshall of old, so have a soft spot for the newcastle
connections (.../mounthost/path/to/file) syntax. I always liked this.
On Wed, Aug 29, 2018 at 9:30 AM Dave Horsfall <dave(a)horsfall.org> wrote:
I know that I'll probably get into trouble for writing this (I'm a born
stirrer, coming from a military family), but...
On Tue, 28 Aug 2018, Kevin Bowling wrote:
This is all so asinine it sounds like it’s run by
a random haughty
social club of people far detached from systems and UNIX and especially
far from publishing systems papers that actually advance the art.
Sometimes it is necessary to let an org know it’s become a joke. I’d
prefer to attend and assist with something not affiliated with Usenix
after reading this thread.
I'm having trouble thinking of an OS (still in current use) that has
remained *mostly* unchanged over 50 years. Yes, Unix has certainly
evolved, but is still recognisable as "do one thing, and do it well".
It's just a shame that they seem to have broken the other philosophy of
"everything looks like a file".