[openbeosnetteam] Re: Haiku internship

  • From: Waldemar Kornewald <wkornew@xxxxxxx>
  • To: openbeosnetteam@xxxxxxxxxxxxx
  • Date: Thu, 01 Jun 2006 22:36:20 +0200

Oliver Tappe wrote:

My name is Andrew Galante, and I am the new intern working on the Haiku
network stack with you this summer.  From talking with Michael, it seemed i
would be working on putting the FreeBSD network stack into Haiku.

Right now, I'm working on getting the 3 OS's (FreeBSD, BeOS R5 and Haiku) up
and running on either real hardware or virtual machines.  I will also be
going through some of the mailing list archives to see what's been going on.

When you check the archives of this list, you'll find that we haven't really made up our minds about which way to go: fix the current stack or replace it with the one from FreeBSD. I am currently trying to find out just how buggy our current stack is and try to get an estimate of how much work would be necessary to bring it into a working state.

I'm pretty sure that we'll port the FreeBSD netstack.

I do not know what Michael (and the others?) has/have told you, but I
personally would find it very helpful if someone (aka: you >;o) would have a
close look at FreeBSD's stack and produce a list of kernel & userland
services that are required by that netstack. You probably know this, but
anyway: the idea was to establish some kind of FreeBSD compatibility "layer"
which would allow their netstack to run on Haiku and still be able to sync
the stack with the FreeBSD repository at a later stage.

Yup, agreed. That would be the first task. I can already name two things:
* sysctl() is not supported by our netstack and the plan is to replace it with a
nicer mechanism (BMessage, and from kernel: KMessage).
* PTHREAD_MUTEX_INITIALIZER is a function call, thus you can't statically
initialize a mutex. ATM, our pthread support is not as nice as it should be. I
know that Axel has the fix on his TODO list, but he's probably busy with other
coding tasks, right now. Anyway, you can get around that even without the fix.


I personally would find it very useful if you and me could collect as much
information as possible about these two different options (stick with our
stack or go FreeBSD) in order to make it easier to decide for the best way to
proceed.

Let me begin. :) The current netstack is: * ugly * buggy (race conditions, etc.) * insecure * not well-designed * not feature-complete * requires a lot of work It basically needs a complete redesign and code review.

Bye,
Waldemar



Other related posts: