Towards an Active Network Architecture

1
Towards an Active Network Architecture Critique

• Rejaie Johnson
• Gergely Biczok

2
Code

• Proposal talks about how we are able to execute
  code and routers
• Potential attacks on ASs, etc.
• What will happen when code has bugs, runs for too
  long or the size of code is too big?

3
Code

• Safety, security, accountability/authentication
  weak arguments given,
• What about non trusted and unworthy code?
• Third parties will they be reliable?

4
Caching and Routing

• Routing caching will be an issue
• ANTS wants to use caching and load code at
  routers.
• Many routers already have their hands full. Do we
  have to put computer farms at every router to
  accommodate capsules?

5
Funding

• Funded September 1996 - August 2000
• This is 2008 What happened?
• Most of the people in the project dont have
  working websites.
• Dont care about your own project?

6
Other issues

• End-End encryption, which is stated
• Sharing State
• What about ISP changing to this? Will this be
  beneficial to them?

7
Program encoding 1.

• Source code
• Human readable and easy to write
• BUT interpreter is needed!
• Significant performance degradation
• Intermediate code
• Java-like
• Wonderful properties
• BUT resource-hogging!
• How much RAM is enough

8
Program encoding 2.

• Platform-dependent code
• Efficient
• Fastest possible
• BUT do you really think that a capsule will
  carry 5-10 different realization of a program
  the actual data???
• 1K of data 550K for the program -gt much ado
  about nothing

9
Interoperability

• shared understanding on resources and their
  names
• BW, CPU, RAM, Disk
• BUT what if new resources are emerging
• Traditional IP networking doesnt have a problem
  with that!

10
Resource safety

• Authorization
• You have to validate all user requests
• BUT how?
• we assume that cryptography will provide the
  basis for the validation mechanism REALLY???
• Overhead potentially on all capsules
• Generating new security problems in an already
  stressed environment
• Without solving any
• Automated delegation of authorization
• Even authors admit this is shaky ground

11
Closing

• IP hourglass -gt agreed program encoding and
  computation environment hourglass
• Destroying the end-to-end principle
• No means for checking if I get the data that I
  asked for (forget about Network Neutrality)
• Interesting idea BUT
• Mostly questions without answers
• Detailed arguments are not convincing