Socket-based transports: Addresses as URIs #7

carletes · 2014-07-06T15:21:12Z

The IPv6 loopback address in my Linux box is usable:

$ /sbin/ifconfig lo
lo        Link encap:Local Loopback  
          inet addr:127.0.0.1  Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING  MTU:65536  Metric:1
          RX packets:689459 errors:0 dropped:0 overruns:0 frame:0
          TX packets:689459 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:43568922 (43.5 MB)  TX bytes:43568922 (43.5 MB)

$ ping6 ::1
PING ::1(::1) 56 data bytes
64 bytes from ::1: icmp_seq=1 ttl=64 time=0.012 ms
64 bytes from ::1: icmp_seq=2 ttl=64 time=0.011 ms
^C
--- ::1 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 999ms
rtt min/avg/max/mdev =0.011/0.011/0.012/0.003 ms
$

My /etc/hosts file, however, does not resolve localhost to an IPv6 address:

127.0.0.1       localhost
127.0.1.1       marcelino

# The following lines are desirable for IPv6 capable hosts
::1     ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters

In this setup the current implementation of whenIPv6 fails to activate the IPv6 tests, because it assumes that localhost must resolve to an IPv6 address in order for IPv6 to be supported:

whenIPv6 assn = do
    addresses <- lookupAddresses NS.AF_INET6 NS.Stream "localhost:1"

My initial workaround was to pass ::1 to lookupAddresses, which would resolve to an IPv6 address, regardless of what localhost resolves to:

whenIPv6 assn = do
    addresses <- lookupAddresses NS.AF_INET6 NS.Stream "[::1]:1"

Unfortunately the current implementation of parseSocketAddress in Network.Transport.Sockets assumes a single colon (if any at all) separating the host name from the port, so the snippet above does not work.

Instead of rewriting the address parsing code in parseSocketAddress, I've sketched this pull requests, which assumes that Address values are URIs, instead of host:port pairs. The version of parseSocketAddress delegates all the parsing work to Network.URI.parseURI.

In this branch I can now run both IPv4 and IPv6 tests:

$ sudo sysctl net.ipv6.conf.all.disable_ipv6=0
net.ipv6.conf.all.disable_ipv6 = 0
$

$ cabal build && ./dist/build/test-courier/test-courier
Building courier-0.1.0.15...
Preprocessing library courier-0.1.0.15...
In-place registering courier-0.1.0.15...
Preprocessing test suite 'test-courier' for courier-0.1.0.15...
hunit: [OK]
endpoints: [OK]
mbox-creation: [OK]
mbox-simple-write: [OK]
mbox-write-read: [OK]
mbox-match: [OK]
mbox-find: [OK]
mem-endpoints+transport: [OK]
mem-bind: [OK]
mem-unbind: [OK]
mem-sendReceive: [OK]
mem-transport: [OK]
tcp-endpoints+transport: [OK]
tcp-bind-unbind: [OK]
tcp-send-receive: [OK]
tcp-double-send-receive: [OK]
tcp-send-receive-reply: [OK]
tcp-multiple-send-receive-reply: [OK]
tcp-local-send-receive-reply: [OK]
tcp6-endpoints+transport: [OK]
tcp6-bind-unbind: [OK]
tcp6-send-receive: [OK]
tcp6-double-send-receive: [OK]
tcp6-send-receive-reply: [OK]
tcp6-multiple-send-receive-reply: [OK]
tcp6-local-send-receive-reply: [OK]
udp-endpoints+transport: [OK]
udp-bind-unbind: [OK]
udp-send-receive: [OK]
udp-double-send-receive: [OK]
udp-send-receive-reply: [OK]
udp-multiple-send-receive-reply: [OK]
udp-local-send-receive-reply: [OK]
udp6-endpoints+transport: [OK]
udp6-bind-unbind: [OK]
udp6-send-receive: [OK]
udp6-double-send-receive: [OK]
udp6-send-receive-reply: [OK]
udp6-multiple-send-receive-reply: [OK]
udp6-local-send-receive-reply: [OK]
call-one-hear-call: [OK]
call-one-call-hear: [OK]
call-concurrent-call-hear: [OK]
call-one-handler: [OK]
call-two-handlers: [OK]
gcall-three-handlers: [OK]
call-one-with-timeout: [OK]
gcall-three-handlers-with-timeout: [OK]

         Test Cases   Total       
 Passed  48           48          
 Failed  0            0           
 Total   48           48          
$

IPv4-only tests seems to work, too:

$ sudo sysctl net.ipv6.conf.all.disable_ipv6=1
net.ipv6.conf.all.disable_ipv6 = 1
$

$ cabal build && ./dist/build/test-courier/test-courier
Building courier-0.1.0.15...
Preprocessing library courier-0.1.0.15...
In-place registering courier-0.1.0.15...
Preprocessing test suite 'test-courier' for courier-0.1.0.15...
hunit: [OK]
endpoints: [OK]
mbox-creation: [OK]
mbox-simple-write: [OK]
mbox-write-read: [OK]
mbox-match: [OK]
mbox-find: [OK]
mem-endpoints+transport: [OK]
mem-bind: [OK]
mem-unbind: [OK]
mem-sendReceive: [OK]
mem-transport: [OK]
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
tcp-endpoints+transport: [OK]
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
tcp-bind-unbind: [OK]
tcp-send-receive: [OK]
tcp-double-send-receive: [OK]
tcp-send-receive-reply: [OK]
tcp-multiple-send-receive-reply: [OK]
tcp-local-send-receive-reply: [OK]
tcp6-endpoints+transport: [OK]
tcp6-bind-unbind: [OK]
tcp6-send-receive: [OK]
tcp6-double-send-receive: [OK]
tcp6-send-receive-reply: [OK]
tcp6-multiple-send-receive-reply: [OK]
tcp6-local-send-receive-reply: [OK]
udp-endpoints+transport: [OK]
udp-bind-unbind: [OK]
udp-send-receive: [OK]
udp-double-send-receive: [OK]
udp-send-receive-reply: [OK]
udp-multiple-send-receive-reply: [OK]
udp-local-send-receive-reply: [OK]
udp6-endpoints+transport: [OK]
udp6-bind-unbind: [OK]
udp6-send-receive: [OK]
udp6-double-send-receive: [OK]
udp6-send-receive-reply: [OK]
udp6-multiple-send-receive-reply: [OK]
udp6-local-send-receive-reply: [OK]
call-one-hear-call: [OK]
call-one-call-hear: [OK]
call-concurrent-call-hear: [OK]
call-one-handler: [OK]
call-two-handlers: [OK]
gcall-three-handlers: [OK]
call-one-with-timeout: [OK]
gcall-three-handlers-with-timeout: [OK]

         Test Cases   Total       
 Passed  48           48          
 Failed  0            0           
 Total   48           48          
$

hargettp · 2014-07-06T15:45:03Z

Good stuff!

What do you think about only parsing a URI directly in the implementation of whenIPv6?

While I can definitely acknowledge that parseSocketAddress is brain-dead (it got the job done through the early phases of writing the code), I'm not sure yet I'm ready to treat all addresses as URIs.

If we just change the definition of whenIPv6 to use a URI to help detect IPv6, then we accomplish the same thing (better detection) with a much smaller change to the overall source.

I have thought about using URIs for addresses, and even thought that might help with deciding which transport should be used for a particular Address. That's why there is a Scheme type that isn't utilized much: if used, it would match the scheme portion of a URI (e.g., http, https, tcp, udp, etc.).

Thoughts?

On Jul 6, 2014, at 11:21 AM, Carlos Valiente [email protected] wrote:

"[::1]:1"

carletes · 2014-07-06T16:26:25Z

What do you think about only parsing a URI directly in the
implementation of whenIPv6?
[..]
If we just change the definition of whenIPv6 to use a URI to help
detect IPv6, then we accomplish the same thing (better detection) with
a much smaller change to the overall source.

Yes, that would also do. That was in fact the way I started toying
around with URIs in order to run the IPv6 tests in my computer. The
main reason why I went for URIs everywhere is that, as you say ...

I have thought about using URIs for addresses, and even thought that
might help with deciding which transport should be used for a
particular Address.

... some bits of Network.Transport.Sockets (lookupAddress & Co) and
Network.Transport.{TCP,UDP} might get simpler after the change to
URIs.

So up to you to decide, I'd say! I'm using this project mainly in order
to
learn Haskell, so I have no strong views on this issue, and I'm happy to
help in either way :)

hargettp · 2014-07-06T16:32:36Z

Well, thanks for the interest!

I think just parsing URIs in the whenIPv6 function would be fine for now; we can save the full URI treatment of Addresses for another day. I'd rather not go all the way treating Addresses as URIs until there has been time to think through the implications.

Thank you!

On Jul 6, 2014, at 11:21 AM, Carlos Valiente [email protected] wrote:

The IPv6 loopback address in my Linux box is usable:

$ /sbin/ifconfig lo
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
UP LOOPBACK RUNNING MTU:65536 Metric:1
RX packets:689459 errors:0 dropped:0 overruns:0 frame:0
TX packets:689459 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:43568922 (43.5 MB) TX bytes:43568922 (43.5 MB)

$ ping6 ::1
PING ::1(::1) 56 data bytes
64 bytes from ::1: icmp_seq=1 ttl=64 time=0.012 ms
64 bytes from ::1: icmp_seq=2 ttl=64 time=0.011 ms
^C
--- ::1 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 999ms
rtt min/avg/max/mdev =0.011/0.011/0.012/0.003 ms
$
My /etc/hosts file, however, does not resolve localhost to an IPv6 address:

127.0.0.1 localhost
127.0.1.1 marcelino

The following lines are desirable for IPv6 capable hosts

::1 ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters
In this setup the current implementation of whenIPv6 fails to activate the IPv6 tests, because it assumes that localhost must resolve to an IPv6 address in order for IPv6 to be supported:

whenIPv6 assn = do
addresses <- lookupAddresses NS.AF_INET6 NS.Stream "localhost:1"
My initial workaround was to pass ::1 to lookupAddresses, which would resolve to an IPv6 address, regardless of what localhost resolves to:

whenIPv6 assn = do
addresses <- lookupAddresses NS.AF_INET6 NS.Stream "[::1]:1"
Unfortunately the current implementation of parseSocketAddress in Network.Transport.Sockets assumes a single colon (if any at all) separating the host name from the port, so the snippet above does not work.

Instead of rewriting the address parsing code in parseSocketAddress, I've sketched this pull requests, which assumes that Address values are URIs, instead of host:port pairs. The version of parseSocketAddress delegates all the parsing work to Network.URI.parseURI.

In this branch I can now run both IPv4 and IPv6 tests:

$ sudo sysctl net.ipv6.conf.all.disable_ipv6=0
net.ipv6.conf.all.disable_ipv6 = 0
$

$ cabal build && ./dist/build/test-courier/test-courier
Building courier-0.1.0.15...
Preprocessing library courier-0.1.0.15...
In-place registering courier-0.1.0.15...
Preprocessing test suite 'test-courier' for courier-0.1.0.15...
hunit: [OK]
endpoints: [OK]
mbox-creation: [OK]
mbox-simple-write: [OK]
mbox-write-read: [OK]
mbox-match: [OK]
mbox-find: [OK]
mem-endpoints+transport: [OK]
mem-bind: [OK]
mem-unbind: [OK]
mem-sendReceive: [OK]
mem-transport: [OK]
tcp-endpoints+transport: [OK]
tcp-bind-unbind: [OK]
tcp-send-receive: [OK]
tcp-double-send-receive: [OK]
tcp-send-receive-reply: [OK]
tcp-multiple-send-receive-reply: [OK]
tcp-local-send-receive-reply: [OK]
tcp6-endpoints+transport: [OK]
tcp6-bind-unbind: [OK]
tcp6-send-receive: [OK]
tcp6-double-send-receive: [OK]
tcp6-send-receive-reply: [OK]
tcp6-multiple-send-receive-reply: [OK]
tcp6-local-send-receive-reply: [OK]
udp-endpoints+transport: [OK]
udp-bind-unbind: [OK]
udp-send-receive: [OK]
udp-double-send-receive: [OK]
udp-send-receive-reply: [OK]
udp-multiple-send-receive-reply: [OK]
udp-local-send-receive-reply: [OK]
udp6-endpoints+transport: [OK]
udp6-bind-unbind: [OK]
udp6-send-receive: [OK]
udp6-double-send-receive: [OK]
udp6-send-receive-reply: [OK]
udp6-multiple-send-receive-reply: [OK]
udp6-local-send-receive-reply: [OK]
call-one-hear-call: [OK]
call-one-call-hear: [OK]
call-concurrent-call-hear: [OK]
call-one-handler: [OK]
call-two-handlers: [OK]
gcall-three-handlers: [OK]
call-one-with-timeout: [OK]
gcall-three-handlers-with-timeout: [OK]
     Test Cases   Total       
Passed 48 48
Failed 0 0
Total 48 48
$
IPv4-only tests seems to work, too:

$ sudo sysctl net.ipv6.conf.all.disable_ipv6=1
net.ipv6.conf.all.disable_ipv6 = 1
$

$ cabal build && ./dist/build/test-courier/test-courier
Building courier-0.1.0.15...
Preprocessing library courier-0.1.0.15...
In-place registering courier-0.1.0.15...
Preprocessing test suite 'test-courier' for courier-0.1.0.15...
hunit: [OK]
endpoints: [OK]
mbox-creation: [OK]
mbox-simple-write: [OK]
mbox-write-read: [OK]
mbox-match: [OK]
mbox-find: [OK]
mem-endpoints+transport: [OK]
mem-bind: [OK]
mem-unbind: [OK]
mem-sendReceive: [OK]
mem-transport: [OK]
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
tcp-endpoints+transport: [OK]
2014-07-06 16:16:46 BST [WARNING] - IPv6 not available
tcp-bind-unbind: [OK]
tcp-send-receive: [OK]
tcp-double-send-receive: [OK]
tcp-send-receive-reply: [OK]
tcp-multiple-send-receive-reply: [OK]
tcp-local-send-receive-reply: [OK]
tcp6-endpoints+transport: [OK]
tcp6-bind-unbind: [OK]
tcp6-send-receive: [OK]
tcp6-double-send-receive: [OK]
tcp6-send-receive-reply: [OK]
tcp6-multiple-send-receive-reply: [OK]
tcp6-local-send-receive-reply: [OK]
udp-endpoints+transport: [OK]
udp-bind-unbind: [OK]
udp-send-receive: [OK]
udp-double-send-receive: [OK]
udp-send-receive-reply: [OK]
udp-multiple-send-receive-reply: [OK]
udp-local-send-receive-reply: [OK]
udp6-endpoints+transport: [OK]
udp6-bind-unbind: [OK]
udp6-send-receive: [OK]
udp6-double-send-receive: [OK]
udp6-send-receive-reply: [OK]
udp6-multiple-send-receive-reply: [OK]
udp6-local-send-receive-reply: [OK]
call-one-hear-call: [OK]
call-one-call-hear: [OK]
call-concurrent-call-hear: [OK]
call-one-handler: [OK]
call-two-handlers: [OK]
gcall-three-handlers: [OK]
call-one-with-timeout: [OK]
gcall-three-handlers-with-timeout: [OK]
     Test Cases   Total       
Passed 48 48
Failed 0 0
Total 48 48
$
You can merge this Pull Request by running

git pull https://github.com/carletes/courier addresses-as-uris
Or view, comment on, or merge it at:

#7

Commit Summary

Socket-based transports: Expect Address parameters to be URIs
Test suite: Improve detection of IPv6 support
File Changes

M README.md (4)
M examples/HelloWorld.hs (6)
M src/Network/Endpoints.hs (4)
M src/Network/Transport/Sockets.hs (55)
M tests/TestTransports.hs (5)
M tests/TestUtils.hs (8)
Patch Links:

https://github.com/hargettp/courier/pull/7.patch
https://github.com/hargettp/courier/pull/7.diff
—
Reply to this email directly or view it on GitHub.

carletes added 2 commits July 6, 2014 15:46

Socket-based transports: Expect Address parameters to be URIs

d9643a3

Test suite: Improve detection of IPv6 support

b41fae0

hargettp added the enhancement label Mar 10, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Socket-based transports: Addresses as URIs #7

Socket-based transports: Addresses as URIs #7

Uh oh!

carletes commented Jul 6, 2014

Uh oh!

hargettp commented Jul 6, 2014

Uh oh!

carletes commented Jul 6, 2014

Uh oh!

hargettp commented Jul 6, 2014

The following lines are desirable for IPv6 capable hosts

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Socket-based transports: Addresses as URIs #7

Are you sure you want to change the base?

Socket-based transports: Addresses as URIs #7

Uh oh!

Conversation

carletes commented Jul 6, 2014

Uh oh!

hargettp commented Jul 6, 2014

Uh oh!

carletes commented Jul 6, 2014

Uh oh!

hargettp commented Jul 6, 2014

The following lines are desirable for IPv6 capable hosts

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants