Re: can anybody review my patchset?

On Mon, Apr 20, 2009 at 3:51 AM, Matthew Dempsky <[email protected]> wrote:
> On Mon, Apr 20, 2009 at 12:29 AM, Andreas Schulze
> <[email protected]> wrote:
>> 6. dnscache rotating A-Records http://thomas.mangin.com/data/source/djbdns-round-robin.patch.bz2

That does seem to be a popular feature.

> Ignoring that, dnscache currently serves recordsets in the same order
> that servers send to it.  I don't see any harm in dnscache permuting
> the order of A records within a recordset, but if you want this sort
> of functionality I would instead do the permutation in query.c rather
> than cache.c like below.  (Patch below just uses dns_sortip.)
>
> (I still don't care for this patch either, because I don't like the
> idea of the client mutating the data already in the cache, but
> cache_get returns a "char *" instead of a "const char *", so I guess
> this is okay...)
>
> --- query.c.orig        2009-04-20 01:46:12.000000000 -0700
> +++ query.c     2009-04-20 01:47:01.000000000 -0700
> @@ -337,6 +337,7 @@ static int doit(struct query *z,int stat
>          goto LOWERLEVEL;
>        }
>
> +       dns_sortip(cached,cachedlen);
>        log_cachedanswer(d,DNS_T_A);
>        if (!rqa(z)) goto DIE;
>        while (cachedlen >= 4) {
>

I've seen variations of the round-robin patch that trigger off of an
environment variable.  Thoughts?

This is one of those features that I'm not sure I personally care much
for, but I've had people tell me it's a must-have for them.  I'm
considering it for zinq-djbdns-0.07.