On 7/30/13 12:29 PM, Brian E Carpenter wrote:
Users want applications to just work, but they (and many business
managers in our "industry") don't understand that when applications
fail unpredictably, it's often because of glitches in what we call
transparency.
I suspect applications are not failing unpredictably often enough,
seriously enough, or impossible-to-fix-edly enough (thank you, I'm
here all week) to provide incentive for radical change in the
network, which elimination of problematic middleboxes would
certainly represent. There are several things going on now
that I think hold some promise for overall approaches to improving
the situation but it's just hard to work on middlebox problems
in the IETF because you can't avoid getting bogged down in rehashing
the same ideological debates over and over and over again. Yes,
problematic middlebox behavior causes real-world problems but
cursing the darkness hasn't yet been productive and I'm reasonably
sure that it never will be, particularly in a world where vendors
are churning out those cursed boxes and network operators and data
centers are buying them.
However, we are in an arms race here. Every step to improve transparency
will be met by a further step in middleboxes that nibbles away at
transparency. We've been debating this for 15 years; have you seen
any real change in the balance of power?
I'm not sure there is a balance of power between the people selling
middleboxes and the IETF. In fact, I'm rather certain there isn't.
Melinda