[dns-operations] Who Ignores TTLs ?

Colm MacCárthaigh colm at stdlib.net
Tue Feb 22 04:01:25 UTC 2011

On Mon, Feb 21, 2011 at 5:19 PM, Paul Vixie <vixie at isc.org> wrote:
> noting that tcpct (RFC 6013) offers the possibility of additional
> savings in the important "time to first eyeball" metric, even greater
> than the best possible results from any conceivable non-tcpct cdn
> technology.  i hope this means our grandchildren will have a cleaner
> design to work with.

That seems like a particularly strange statement, it's hard to see how
tcpct has any impact.

On the time-scale you mention though; one conceivable change is that
as network transmission and server capacity become cheaper and cheaper
commodities, a tipping point will come when the best architecture will
be for eyeball-side endpoints to communicate with all possible server
endpoints concurrently.

The least-latent response can be used where interaction matters,
consensus can be checked where coherency and security matter, and
faults tolerated without the need for aggressive retries.

In other words; when the network is so cheap it no longer seems like
waste, client software will probably start to resemble today's
advanced recursive resolvers.


More information about the dns-operations mailing list