[openbeosnetteam] Re: Tests plan?


On Jan 11, 2004, at 4:54 AM, Philippe Houdoin wrote:

Hey Gang,

:-)

I also think we should capture a core dump on Philippe's and
Waldemar's
brain.  You and I are coming into this cold (which has its
advantages),
but a quick proverbial "state of the union address" definitely
couldn't
hurt.


Just wait my next brain crash, and I'll post you the core file :-p
Should not take too much time, as I'm under heavy pressure at
University these weeks,
a segfault is higly expected shortly...

:-D

I don't how best we, I and Waldemar, could hands you a good, detailed
picture of current stack inner working.
Guess it's time we add content to this little current/docs/develop/net/
stack.html page!

Yeah, an architectural diagram would be most helpful. Brennan has a good start on that already.


And/or you can ask specific questions, we'll try our best to answer
them.

I personally am testing the stack initially by running through each
of our available network applications:

arp
ifconfig
ping
route
traceroute


How about more abusive things like nmap (or the Be equivalent), ping
floods, and multicast?

Do we have a way to test things like IGMP or PIM? What about UDP? 

Testing IGMP is easy: we don't have any support yet!
But maybe you mean ICMP here, right?

Nope, IGMP. "Internet Group Management Protocol." The main reason I suggested it is because of all the multicast protocols this one is relatively painless to use as a test for both the stack and ethernet drivers' multicast handling.

Anyway, I agree, we need some reliable TCP/IP testing
tool. Maybe we can port an open-source one available in the Web
wildness?
Who wants to hunt it?

I'll do it. I'm thinking netperf which besides abusing the stack also gives us real performance statistics, and allows us to test TCP and/or UDP. Netperf has been around for a long time and is a pure U*ix program (hopefully SCO won't try to claim ownership). I'm guessing that porting efforts shouldn't be overly burdensome.

Cheers,
Scott


Other related posts: