IPv6 support

I started implementing some bits of IPv6 support, namely tango.net.Socket.IPv6Address, and have a design question. The usual way to create an Address from a host name is new IPv4Address("dsource.org"). This means that any application won't work with IPv6 without some code changes, which may break IPv4 compatibility. So, instead of directly creating IPv4Address or IPv6Address instances, there should be a factory method, possibly Address Address.get(char[] host) and variations. The issue is that you'd either need some registration mechanism for the address families, or have Address depend on IPv4Address, IPv6Address and others. I'd like to hear your opinion on that.

Why not add a new module HostResolver?, or something, to move the dependency away from Address?

I really like that suggestion. More opinions, anyone?

Ok. Sorry for sleeping. I had to change quite a few things because I removed IPv4Address, moved all the functionality (IPv4 and IPv6) into InternetAddress and that class into tango.net.Socket, with nothing more than a public import in tango.net.InternetAddress. Of course, I am open for suggestions how to do this more cleanly, but I find this solution simple and functional.

I have attached a rather mighty patch for revision 3936. Please give it some testing. Much testing. Both for IPv4 and IPv6. I'm especially concerned about IPv4 applications breaking. Thanks!

Does this affect what is asked for in #1300 ?

anyone done further testing on this? I've been swamped, with no time to get to it yet ...

No, name and nameLen are still protected. Are there any reasons for not just making those public?

I superficially tested UDP on Linux, and it seems to work fine. Maybe I broke some broadcast or multicast functionality, or introduced unnecessary source incompatibilities. Name some complex application or library and I will test it.

I'll give Windows XP some love and implement the missing bits within the next week or so.

I suppose the only reason to suppress them is that the structure of that module may change over time, that's about all I can think of.

I'll test the XP stuff when it's ready :)

elite01: Please see #1228

There is a typo in the unitest of the patch: IneternetAddress?

elite01: did you get a chance to revise this since we last spoke?

I've upload the patch, please test it and say, what is need else?

Thank you, we will

I've update the IPv6.patch. Now it works under 32 bit Linux. I hope it will be works under other 32 bit posix OS. Please test it under Windows and 64 bit posix.

thanks! Would you mind adding a little bit of documentation explaining what those new functions do? We're not sufficiently familiar with IPv6 to grok their usage right away :)

And some changes. Also added a few comments. What aspects need to be documented?

Replying to Wazar:

And some changes. Also added a few comments. What aspects need to be documented?

The new functions (mostly done already), and perhaps a little description (at the top) about how IPv6 differs in usage from IPv4? Nothing fancy ... perhaps a link to some online documentation if you know of a good explanation?

Thanks!

I added comments at the top of IPv6Address class declaration.

will merge this during the week - thanks!

(In [5409]) fixes #1239 :: IPv6 support

Big thanks to Wazar

doesn't exist on Windows: gai_strerror(int errcode);

Can you find a resolution, please? :)