On Sat, 13 Mar 2010, John M McIntosh wrote:

> Ok, well I spent the evening looking into this. It's unclear if reverting the new/old socket stuff in Pharo is a good idea,
> or if we just adjust things a bit in the name lookup that would solve things. There are two questions to resolve
>
> (1) Do we want to return an IPV4 address from the network lookup? or an IPV6 address?
> (2) What do we do when we have 2 or more active IP interfaces on the machine, prompt for which one to use? Pick one at random, use the first or last one?

What about creating a new primitive which would return a list with ip 
address - interface pairs. The list would contain all ip addresses so we 
can solve the problem in smalltalk. If we later find that the initial 
solution was wrong, we can change it without touching the vm code.


Levente

>
> If the community can decide what to do, then we can propose a solution.
>
> So how it works.
>
> NetNameResolver localHostAddress
>
> fe80::21c:42ff:fe00:9%en5(otter-2.local),0(0)
>
> Gah... what's that?
>
> Well the interface is:
>
> en5: flags=8963<UP,BROADCAST,SMART,RUNNING,PROMISC,SIMPLEX,MULTICAST> mtu 1500
> 	inet6 fe80::21c:42ff:fe00:9%en5 prefixlen 64 scopeid 0x8
> 	inet 10.37.129.2 netmask 0xffffff00 broadcast 10.37.129.255
> 	ether 00:1c:42:00:00:09
> 	media: autoselect status: active
> 	supported media: autoselect
>
> Thus it's giving me an IPV6 address.
>
> The chore is more complicated that one would care to see. I see Microsoft had a hand in defining the spec.
>
> So it first asks the gethostname for the name of the host.  As an example my macbook pro has
> Otter-2.local
>
> which is what
> [otter-2:~] johnmci% hostname
> Otter-2.local
>
> That's a bonjour assigned name since if I do
> [otter-2:~] johnmci% nslookup Otter-2.local
> Server:		192.168.1.7
> Address:	192.168.1.7#53
>
> ** server can't find Otter-2.local: NXDOMAIN
>
>
> Then we are off to:
>
> getnameinfo
>
> which returns a chain of  IPV4 and IPV6  address that Otter-2.local would resolve to.
> In this case there are six address, broken in to IPV6 and IPV4
>
> fe80::21c:42ff:fe00:8%en4(otter-2.local),0(0)-inet6-stream-tcp
> fe80::21c:42ff:fe00:9%en5(otter-2.local),0(0)-inet6-stream-tcp
> fe80::21b:63ff:fe02:d2db%en1(otter-2.local),0(0)-inet6-stream-tcp
> 10.211.55.2(10.211.55.2),0(0)-inet4-stream-tcp
> 10.37.129.2(10.37.129.2),0(0)-inet4-stream-tcp
> 192.168.1.141(192.168.1.141),0(0)-inet4-stream-tcp
>
> So looks kinda random, but it's not. It's defined by Microsoft & committee etc. in
> http://www.faqs.org/rfcs/rfc3484.html
>
> The objective is to sort the list into some order. I must say you can't actually tell from reading the docs, and things over
> the past decade have changed. So I refer to some cheat sheets that *might* be correct.
>
> Now the macbook pro in this case uses 192.168.1.141 as the assigned tcp/ip address from our internal DHCP server.
> The 10.211.55.2  is a parallels shared network adaptor and
> 10.37.129.2 is the parallels host-only network adaptor.
>
> All three interfaces show as active, and if you consider the 'ifconfig -a'  below you would be hard pressed to determine which interface is the one facing the
> company intranet.
>
>
> [otter-2:~] johnmci% ifconfig -a
> lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 16384
> 	inet6 ::1 prefixlen 128
> 	inet6 fe80::1%lo0 prefixlen 64 scopeid 0x1
> 	inet 127.0.0.1 netmask 0xff000000
> gif0: flags=8010<POINTOPOINT,MULTICAST> mtu 1280
> stf0: flags=0<> mtu 1280
> en0: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> 	ether 00:17:f2:d9:57:35
> 	media: autoselect status: inactive
> 	supported media: autoselect 10baseT/UTP <half-duplex> 10baseT/UTP <full-duplex> 10baseT/UTP <full-duplex,hw-loopback> 10baseT/UTP <full-duplex,flow-control> 100baseTX <half-duplex> 100baseTX <full-duplex> 100baseTX <full-duplex,hw-loopback> 100baseTX <full-duplex,flow-control> 1000baseT <full-duplex> 1000baseT <full-duplex,hw-loopback> 1000baseT <full-duplex,flow-control> none
> fw0: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 2030
> 	lladdr 00:19:e3:ff:fe:93:92:7c
> 	media: autoselect <full-duplex> status: inactive
> 	supported media: autoselect <full-duplex>
> en1: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> 	inet6 fe80::21b:63ff:fe02:d2db%en1 prefixlen 64 scopeid 0x6
> 	inet 192.168.1.141 netmask 0xffffff00 broadcast 192.168.1.255
> 	ether 00:1b:63:02:d2:db
> 	media: autoselect status: active
> 	supported media: autoselect
> en4: flags=8963<UP,BROADCAST,SMART,RUNNING,PROMISC,SIMPLEX,MULTICAST> mtu 1500
> 	inet6 fe80::21c:42ff:fe00:8%en4 prefixlen 64 scopeid 0x7
> 	inet 10.211.55.2 netmask 0xffffff00 broadcast 10.211.55.255
> 	ether 00:1c:42:00:00:08
> 	media: autoselect status: active
> 	supported media: autoselect
> en5: flags=8963<UP,BROADCAST,SMART,RUNNING,PROMISC,SIMPLEX,MULTICAST> mtu 1500
> 	inet6 fe80::21c:42ff:fe00:9%en5 prefixlen 64 scopeid 0x8
> 	inet 10.37.129.2 netmask 0xffffff00 broadcast 10.37.129.255
> 	ether 00:1c:42:00:00:09
> 	media: autoselect status: active
> 	supported media: autoselect
>
>
> Where en0 is ethernet jack, fw0 is firewire, en1 is wireless, en4 & en5 are virtual.
>
> Now the bias from the committee is
>
> precedence  ::1/128       50
> precedence  ::/0          40
> precedence  2002::/16     30
> precedence ::/96          20
> precedence ::ffff:0:0/96  10
> scopev4 ::ffff:169.254.0.0/112  2
> scopev4 ::ffff:127.0.0.0/104    2
> scopev4 ::ffff:10.0.0.0/104     5
> scopev4 ::ffff:172.16.0.0/108   5
> scopev4 ::ffff:192.168.0.0/112  5
> scopev4 ::ffff:0.0.0.0/96       14
>
> This means of course the 10.0.0.0 address sorts before the 192.168.0.0 address. How helpful.
>
> So how do we know *which* is the proper address to use? Well no idea!
>
> However let's see what the API does.
>
> NetNameResolver class>> addressesForName: hostName
> 	| adresses |
> 	adresses := SocketAddressInformation
> 		forHost: hostName
> 		service: ''
> 		flags: 0
> 		addressFamily:  0
> 		socketType: SocketAddressInformation socketTypeStream
> 		protocol: SocketAddressInformation protocolTCP.
> 	^adresses
>
>
> Now I can make it return just IPV4 address by doing:
>
> NetNameResolver class>> addressesForName: hostName
> 	| adresses |
> 	adresses := SocketAddressInformation
> 		forHost: hostName
> 		service: ''
> 		flags: 0
> 		addressFamily:  SocketAddressInformation addressFamilyINET4
> 		socketType: SocketAddressInformation socketTypeStream
> 		protocol: SocketAddressInformation protocolTCP.
> 	^adresses
>
> which then NetNameResolver localHostAddress gives:
> 10.37.129.2(10.37.129.2),0(0)
>
> I can also supply a service number string  say SSH  '22'
> that give back
> 10.37.129.2(10.37.129.2),22(ssh)
> since ssh is binding on all interfaces that's valid.
>
> Technically the getnameinfo gives back three address but again the code below grabs the first one which according to the rfc3484 is more likely to be the correct answer.
>
> addressForName: hostName
> 	"NetNameResolver addressForName: 'impara.de' "
> 	"NetNameResolver addressForName: 'localhost' "
> 	"NetNameResolver addressForName: '127.0.0.1' "
> 	| addresses |
> 	self useOldNetwork
> 		ifTrue: [^self oldAddressForName: hostName].
> 	addresses := self addressesForName: hostName.
> 	^addresses
> 		ifEmpty: [nil]
> 		ifNotEmpty: [addresses first socketAddress]
>
> But grabbing the first one doesn't meet people's expectations of correctness, however we just don't have enough information to decide *what* is correct.
>
> Now some people say OH let's return the en0 one because that is correct? Really how do you know?
> On my computer en0 is ethernet, but since I'm using wireless then en1 is the correct one.
>
> Obviously I could look in the list of IPV6 address find the LOWEST en# value, then grab the IPV4 entry.  This would assume I check for the fact that the IPV6
> interface has an IPV4 address and the sort order for the IPV6 is the same as for the IPV4.
>
> Gee that sounds ok?
>
> Well it's not, when for example I tether my MacBook Pro to my iPhone and abuse my privilege to move data over the cellular carrier, ah sorry you USA ATT folks,
> then the interface is:
>
> en6: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> 	inet6 fe80::226:8ff:fe72:5aa%en6 prefixlen 64 scopeid 0x9
> 	inet 192.168.20.2 netmask 0xffffff00 broadcast 192.168.20.255
> 	ether 00:26:08:72:05:aa
> 	media: 10baseT/UTP status: active
> 	supported media: 10baseT/UTP
>
>
> which fails the test that the lowest active en# is the one we want to use.
>
> 10.37.129.2(10.37.129.2),0(0)-inet4-stream-tcp
> 10.211.55.2(10.211.55.2),0(0)-inet4-stream-tcp
> 192.168.20.2(192.168.20.2),0(0)-inet4-stream-tcp)
>
>
> So the question is what would the community like to see happen. Well other than hide under a rock and see if the problem goes away...
>
>
> On 2010-03-12, at 8:41 AM, Chris Cunnington wrote:
>
>> How is NetNameResolver choosing the interface to return an IP address from?
>>
>> In Cobalt this is a problem. You need the right address to connect to other spaces. Cobalt won't let you enter it by hand, so you are at the mercy of its guessing the correct interface.
>>
>> NetNameResolver localAddressString
>>
>> executed in Workspace produces the IP from my fw0 interface, which is useless. I need it to look at ppp0.
>>
>> Can it be told which interface to look at?
>> How does it choose?
>>
>> Chris
>>
>
> --
> ===========================================================================
> John M. McIntosh <johnmci at smalltalkconsulting.com>   Twitter:  squeaker68882
> Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
> ===========================================================================
>
>
>
>
>