Subject: Re: [PATCH] Return TTL data

Re: [PATCH] Return TTL data

From: Steinar H. Gunderson <sesse_at_google.com>
Date: Fri, 12 Oct 2007 14:48:21 +0200

On Fri, Oct 12, 2007 at 01:52:36PM +0200, Bram Matthys (Syzop) wrote:
> I don't understand, why does this has to be an 'x AND y' question?
> Isn't it possible to threat them differently?
> As I stated I personally see a big difference between breaking the API for
> some timeout statistics in the general resolver routines (which are likely
> to be most used) versus returning the parsed ttl in a *record parse function*.

I guess it's a matter of whether you view breaking the API as a mostly binary
operation or not. Of course you could insert only one patch and not the
other, but once you've already broken the API at one place, breaking it the
other one (in the same c-ares version) won't hurt too much, will it?

> Anyway, I don't expect any change, I justed wanted to state my concerns /
> opinion so you will give this more thought in the future.

I don't expect to be publishing any more API-breaking patches to c-ares. I
agree with your general comment that APIs should not be broken lightly,
though -- I can agree that having timeout values on its own would probably be
a dubious case, but I don't see a big problem in let it "piggyback" on the
the TTL stuff. (This is, of course, logically -- the timeout code is
committed already, while the TTL patches still wait for review here.)

/* Steinar */

-- 
Software Engineer, Google Norway
Received on 2007-10-12