On killing IPv6 transition mechanisms
tore.anderson at redpill-linpro.com
Tue Mar 16 10:13:18 CET 2010
* Ted Mittelstaedt
> Please quit FUDDing. Content providers only care that customers on
> the Internet are able to get at their offerings. Ebay, Craigslist
> and Amazon don't give a crap about latency of a connection if they
> get a credit card number over the connection for a valid order. And
> there are still a VERY LARGE minority of people on dialup-only IPv4,
> so quit being a connectivity snob - the worst IPv6 broadband
> connection out there is going to have lower latency than a 56k dialup
> While the IPv6 deployment isn't served by "broken IPv6 connectivity"
> any content provider looking at IPv6 knows 2 things, first that most
> if not all IPv6 connections to their dual-stacked servers are "beta
> testing", (since IPv4 is still available from the RIR's) second that
> the more connectivity options they have to their servers the greater
> the chance that someone will be able to connect to them and get their
> You seem to assume the content providers on the Internet are idiots.
> They know that IPv6 right now is in it's infancy, and they are
> watching the usage of it. What is more important right now is
> GETTING the IPv6 connections from users, not how good they are.
> That's what they are paying attention to.
it might seem reasonable to assume that adding IPv6 capability to a web
site will make it available for more eyeballs compared to running with
IPv4-only. Unfortunately, the opposite is true. If you add
AAAA-records to your web site, the total number of eyeballs able to
access your content decreases - not by a big number, but a measurable
one. As you correctly point out, this is something content providers
care a great deal about.
However, you're mistaken that content providers do not care about
latency. Actually, what they do care about is providing the best
browsing experience to their users - they will do other stuff to
accomplish that too, such as creating light-weight web pages for mobile
users, write HTML in a way that minimises browser rendering latency, use
different TCP protocol parameters depending on the client are in a
mobile network or not, selectively enable/disable compression in the
application layer depending on the client's connectivity - the list goes
on and on. Network latency is without question on that list. While
nothing can be done about a user's high-latency connection, 150ms of
additional client-server latency due to IPv6 being used will quickly add
up to seconds in terms of overall page load time for a complex web site,
and that's is certainly something users will notice, especially the
broadband users who are expecting things to be snappy.
So when considering dualstacking content, there's two arguments agaist
at the moment: first that it will result in the site being unreachable
for a small set of users, second that it will result in decreased
connection quality for another small set of users. There are no real
arguments for, except for ones like «it's cool». Sadly. However I
think we're soon approaching the point where the number of users
affected in a negative way are small enough that the coolness factor
will trump the disadvantages, at least for content providers where the
techies call the shots.
I'm in the content hosting business, by the way. All of the things I
point out above are real concerns from real content providers.
Redpill Linpro AS - http://www.redpill-linpro.com/
Tel: +47 21 54 41 27
More information about the ipv6-ops