Martin wrote:
Hello Taylor,
...
The "netecho" and "nettest2" tools are unfinished and unmaintained.
Heck, "netecho" cannot even setup the socket correctly if you want to
use it with TCP.
Right, "# netcho -t SOCK_DGRAM -p 8080" + "$ nc -u localhost 8080" will work if you are
patient enough to wait for the first ARP timeout. "# nettest2" will not work since, it appears, we
have some regression in nildummy which dumps core - probably a victim of some cleanup in the lower levels of
the stack :-(
In retrospect, it was probably a bad decision to integrate the current
networking implementation into the mainline as-is, but back then we
really thought that we had to deal with only a few glitches that are
easily fixable. After a while we found out that local fixes won't
salvage the networking, because the code is just plainly wrong. What you
observe are the many ill-effects of its quality. For example, in your
scenario the first packet from netcat gets lost because the ARP server
does not send a proper ARP reply for the first ARP request.
It works somehow, but it is very fragile... and unmaintainable.
Otherwise let us hope that Jiri will be able to maintain his pace and
finish his complete rewrite of the whole networking stack soon. You can
follow his work-in-progress in his branch:
https://code.launchpad.net/~jsvoboda/helenos/inet
Hopefully I would be able to merge another part of my work this or next week.
In my branch UDP should be already on par with the mainline. Plus the annoying
ARP bug is not present.
I plan to start a networking section in the user's guide after I merge my
changes (since this completely changes the way you configure networking).
Cheers
-Jiri
_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/cgi-bin/listinfo/helenos-devel
