gethostbyname() and resolv.conf updates
sgallagh at redhat.com
Fri Jun 18 14:12:34 UTC 2010
On 06/18/2010 09:59 AM, Colin Walters wrote:
> On Fri, Jun 18, 2010 at 9:56 AM, Stephen Gallagher<sgallagh at redhat.com> wrote:
>> Sorry, my reply was more directed against the assertion that NSCD should
>> be mandatory to solve this without changes to glibc. I agree that it
>> would be ideal for gethostbyname() to internally perform a res_init()
>> when appropriate.
> Ok, then we agree. I don't personally care whether it's NSCD or a
> magical oracle as long as our operating system's gethostbyname()
> function actually functions on mobile devices. (Caching DNS would be
> a huge bonus, as other operating systems do this and us not doing so
> makes the internet "feel" significantly, visibly slower compared to
> them, but that's a separate bug)
Well, I definitely care that it's not NSCD, as NSCD is such a broken
caching mechanism that basically everyone turns it off in any
environment where reliable caching is needed (and is also a large part
of the reason SSSD was developed - to produce a stable cache).
If we start asserting that everyone has to use NSCD in order to get
stable gethostbyname(), then clearly someone is missing the point.
Delivering value year after year.
Red Hat ranks #1 in value among software vendors.
More information about the devel