Re: MX lookup problems

[Brian W. <bwe...@xb...>]

On Thu, 5 Sep 2002, Andrew Houghton wrote:

> I saw a post on the sourceforge discussion boards about this same 
> question, followed by a response of Brian's saying the lists were a 
> better spot to ask questions, so here I am..
> 
> We're using dnsjava to do minimal domain validation on email addresses. 
>  The java code looks something like (this is from memory, so forgive any 
> syntax/semantic errors, the idea should be clear):
> 
> boolean checkDomain(String domain) {
>   Record[] records = dns.getAnyRecords(domain, Type.MX);
> 
>   if (records == null) {
>     // no MX records, check for CNAME
>     records = dns.getAnyRecords(domain, Type.CNAME);
> 
>     if (records != null) {
>       // found a CNAME, start over
>       return checkDomain(records[0].rdataToString());
>     }
> 
>     // no CNAME records, check for A
>     records = dns.getAnyRecords(domain, Type.A);
> 
>     if (records == null) {
>       return false;
>     }
>   }
>   
>   return true;
> }
> 
> 
> I think this satisfies the RFC 2821 address discovery protocol.  I have 
> a couple of questions, however:
> 
>    1. it doesn't handle CNAME loops nicely.  I'm willing to hear any
>       suggestions on this; I'll probably simply limit the recursion depth.

dns.getAnyRecords() should already follow CNAMEs.  In fact, there's no way 
to tell dns.getAnyRecords() to not follow CNAMEs.

>    2. it doesn't always work.  For example, dig gives me an MX response
>       for the domain 'phipps.pgh.pa.us', but the dns.getAnyRecords()
>       call doesn't.

That's a good question.  It appears to be a bug somewhere; I'll look into 
it later today.

>    3. it doesn't timeout nicely.  I'd like to set a timeout across all
>       of these queries -- that is, I'd like a way to say that the
>       top-level call to checkDomain() will respond within 2 seconds.
>        Can anyone think of a reasonably simple way of ensuring that?
>        I'd really rather not have to write my own thread pool
>       implementation.

Not really.  You could use a Timer, but there's no way of doing it within 
dnsjava.

>    4. the granularity of the Resolver.setTimeout() method is one second
>       (why?), which makes it hard for me to even fake the 2 second
>       timeout.. with a recursion limit of 1 and the timeout set to it's
>       minimum, I could still end up with a 4 second wait.

The granularity is one second because I didn't think anyone would need 
anything finer.  It would be pretty easy to add a

	setTimeout(int secs, int millisecs);

function to the Resolver interface, if you think that would be useful.


> I'm looking for any suggestions I can get here.. I seem to have 
> identified a bug in the package (point 2, above) but the others are 
> purely design problems.  I'd also like to thank Brian for producing what 
> appears to be a great package -- sure beats having to try and cobble 
> together something like this myself.

I'm glad it's useful.  Thanks!

Brian