I have to agree with Alex, BUFSIZ of 1024 and PIPE_BUF of 512 is very
archaic. Disks are getting bigger and bigger. and 4096 disk sectors are
becoming common, if not already the standard (they may be getting even
bigger - as memory and disks expand).
However: Questions back to Alex: How large are the atomic messages
over pipes in picolisp, on average? I think you don't actually send the
data over the pipe, but just block addresses in a DB have been updated. I
am curious about this myself.
On Aug 25, 2011, at 3:12 PM, Amit Kulkarni wrote:
> Background: I am trying to port picoLisp with Alex's help, its written
> in assembly for 64 bit architectures and C for 32 bit arches. He is
> ported to amd64, ppc64 and sparc64 is on his roadmap.
> Can somebody please weigh in on Alex's comment?
> ---------- Forwarded message ----------
> From: Alexander Burger <x...@software-lab.de>
> Date: Thu, Aug 25, 2011 at 6:00 AM
> Hi Amit,
> I'm surprised to see that OpenBSD uses only 1024 for BUFSIZ, and - even
> more - that PIPE_BUF is only 512 (all others so far use 4096 or 5120).
> Especially the size of PIPE_BUF is critical for performance, as this is
> the limit for atomic messages over pipes, used heavily in PicoLisp's
> interprocess communication.
> Does OpenBSD really need to be so much compatible with ancient Unix?