Murali Vilayannur wrote:
Hey Rob,
If you're going to do TCP addresses, the normal thing to do is
address/netmask pairs. tcp://192.168.1.0/8 would be more or less how
nmap would do it. tcp://192.168.1.0/255.255.255.0 would be one way the
NFS exports file would accept it (the other is valid for older NFS
versions too, I think).
Okay. I think we should be able to change that.
Wouldn't wildcards like tcp://192.168.1.* achieve the same purpose? I must
be missing something...
You can define subnets that use just part of the bits in the quad for
the network. The '*' notation doesn't cover those cases.
function. That's not so great, but I don't think that we should worry
about it for now.
Yeah. if people want all these options, then they pay for them in terms
of performance :)
If no options are specified it will fall thru quickly with no string
comparisons.
Yep!
One other thing I'd like to see is a function or two for pulling out
these lists of addresses in the config parsing code, to keep us from
having N copies of the same stuff.
okay. will refactor the parsing code into a common function.
Thanks!
Last thing: I'm not convinced that we should do anything different for
the Security stuff, other than possibly moving it into the same
ExportOptions section as everything else. We can try to make that more
generic some other time.
Yup. That is what I had in mind (moving it as part of the ExportOptions
section and having a tag like TrustedPorts <list of wildcard BMi-tcp url's/netmask
combinations>
and TrustedNetwork <list of wildcard BMI-tcp url's/netmask combinations>.
Sounds great. Thanks,
Rob
_______________________________________________
PVFS2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
