Re: java.net: Classpath vs. libgcj Comparison

[Tom Tromey <tromey at cygnus dot com>]

>>>>> "Aaron" == Aaron M Renn <arenn@urbanophile.com> writes:

Aaron> I believe there is a free 100% Java implementation of DNS
Aaron> called DNSJava, which would allow a 100% Java implementation of
Aaron> InetAddress.  I have not looked at this package in detail, but
Aaron> the homepage is http://www.xbill.org/dnsjava/.  It is licensed
Aaron> under the LGPL, but the author might allow libstdc++ license.
Aaron> This implementation also contains caching, I noticed.  My only
Aaron> question is how to configure it, which the README file
Aaron> indicates is via a property, with fallback to /etc/resolv.conf
Aaron> and several other sources.  It might be worth a look.  For
Aaron> ideas if nothing else.

One problem with this approach is that it adds another layer of
porting problem: you have to modify it for each system to make it
understand how lookups are done.  This is particularly ugly on systems
that have things like /etc/nsswitch.conf.

It sucks that the C library interface to the resolver is so bad.  If
it were better we could get the TTL and properly implement caching.
(Complaining about this doesn't help, but it makes me feel better.)

On this one I have to side with Jeff Sturm: simple caching is clearly
wrong for long-running applications.

>> One item I didn't see in Gaute's comparison is any mention of IPv6
>> conformance.  That could come in handy.

Aaron> The Classpath implementation does not support IPv6 currently.
 
libgcj's supports it to some degree, but I don't know precisely how
well.  Warren has the state on this -- he wrote java.net.


I'm very happy with Gaute's comparison.  It is good to see progress
being made.  I wish I had time to help with this process.  Hopefully I
will in the not-too-distant future.  We'll see.

Tom