The issue that scares me is “Who do you trust?”
If a new host comes along and says “Hi, My name is Bill. What’s yours?
Do
you want to go out some time?” The other host has no way of knowing if
“Bill” is a trusted new member of the community, or some predator that
snuck
in. At least that’s what all the girls tell me when I try internet
dating.
Somewhere there NEEDS to be a trusted, authoritative list of who’s who.
I guess that a host could contact all of their already trusted friends and
ask if they know “Bill”. If they do and Bill is trusted this host could
just add him to it’s list of trusted friends. But as the trusted network
gets bigger and bigger, this becomes unreasonable pretty fast.
“Kevin Stallard” <> kevin@ffflyingrobots.com> > wrote in message
news:apcnk2$lng$> 1@inn.qnx.com> …
“Kevin Stallard” <> kevin@ffflyingrobots.com> > wrote in message
news:apbri7$lla$> 1@inn.qnx.com> …
Well, for now, yes…but this is something we are working on moving
away
from. This would remove an additional dependancy from the application
if
we
didn’t have to care about node names.
Let me clarify this a little bit.
Node names are still extremely useful. It helps you identify which
physical
piece of hardware to which you are talking. Very important when it
comes
to
diagnostics. Robotics it is especially important. In the robotics
case,
you may want to introduce a new machine w/o having to tell everyone else
its name.
I don’t think a global name locator service is the answer, but the
ability
to discover who is there and what resources they have is. I really
think
you guys hit a home run when you decided to have servers show up in the
pathname space. Yeah…when you query for this info there are a lot of
stat()'s going on, but this should only occur during initial discovery.
I’m
not sure its that big a deal as when you know what services are there,
you
connect to them usuall for the lifetime of what ever operation you are
performing.
Anyone knows that if you are going to access a file, you don’t open and
close, then re-open it for every record.
We just need to be able to broadcast a “tell me who you are” message to
currently running nodes and get responses on demand. The broadcast
idea
will work, but like you said xtang, it is going to result in a lot of
chatter. It would be nice if remote qnets could respond to a “tell my
who
you are” broadcast.
Kevin
Kevin
“Xiaodan Tang” <> xtang@qnx.com> > wrote in message
news:apbrf9$n7s$> 1@nntp.qnx.com> …
Kevin,
On a perticular system, don’t you always KNOW all your node names?
You
know
you can
talk to a node even if it is not show up under /net, do you?
-xtang
Kevin Stallard <> kevin@ffflyingrobots.com> > wrote in message
news:apbp18$ipj$> 1@inn.qnx.com> …
On the other hand, the undocumented “broadcast=” option, would
make
it
chatty, but
somewhat reliable (on node being show up under /net).
Let me point this out, there is NO reliable way to make “all
node
show
up”
from the design
view of QNET. What if QNET is based on tcpip, do you want to
know
ALL
the
nodes
on the otherside of earth is come and go? What if some node is
on
a
flaky
link, and not
broadcast reachable?
The TCPIP part of this is understandable. That would be too hard
to
do,
but
it is unlikley that we would do what we want using TCPIP as it
would
be
local LAN only (TCPIP has too much overhead, and too much has to
be
done
to
make it dynamic (have to use DHCP, etc). This is how we intend to
use
it
(local LAN only)…
Thanks for your feedback xtang…it’s appreciated.
Kevin
This will only work on some special condition (a LAN, reasonable
reliable
media).
-xtang
Kevin
“Xiaodan Tang” <> xtang@qnx.com> > wrote in message
news:apaapi$lvb$> 1@nntp.qnx.com> …
Kevin Stallard <> kevin@ffflyingrobots.com> > wrote in message
news:apa9o0$r0s$> 1@inn.qnx.com> …
Hey xtang…how’s things goin?
I want to be able to start a server on any node and be
able
to
find
it
w/o
knowning the name of the node on which it is residing.
This
would
allow
This is what name_attach(name, NAME_ATTACH_FLAG_GLOBAL)
and name_open(name, NAME_ATTACH_FLAG_GLOBAL) suppose
to do.
Unfortunatly this is not true right now. We will have a
“Global
Name
Services”
manager to address this.
-xtang
dynamic shifting of processes and the dynamic discovery of
them.
I need a way to find out the names of the nodes on the
network
so
I
can
query each one…
Thanks,
Kevin
“Xiaodan Tang” <> xtang@qnx.com> > wrote in message
news:apa9ik$l7b$> 1@nntp.qnx.com> …
No, netmgr_ctl() won’t help. In fact, there is no way in
QNET
to
re-learn
the existing node.
Node names under /net is not that “reliable” anyway.
It might be interesting to know why you want to do this?
and
maybe
there
are
otherways
to do it.
-xtang
Kevin Stallard <> kevin@ffflyingrobots.com> > wrote in
message
news:ap9ieq$337$> 1@inn.qnx.com> …
I found this function mentioned in the QNET Networking
docs,
netmgr_ctl(),
but it doesn’t seem to exist in the library reference.
Would
netmgr_ctl()
help me out with this?
Thanks
Kevin
“Kevin Stallard” <> kevin@ffflyingrobots.com> > wrote in
message
news:ap9gpk$1e2$> 1@inn.qnx.com> …
Hi,
I’d like to be able to send a message to io-net/qnet
to
force
it
to
query
for existing nodes running qnet. I am going to be
looking
for
server
names
by scanning the /net directory, and have noticed
that
unless
I
know
a
name
of a node, I can’t get qnet to display until it
decides
to
take
a
look
and
see who is there.
Is it possible?
Thanks
Kevin
\
