Students who want to read more on computer networking can read:

RFC 3168 (ECN, Sept 2001) Try to read the first few pages! If 
         you get hooked, read the rest.

RFC 1122 (Host Requirements). Sections 3.3.2 and 3.3.3 explain 
         EMTU_R and EMTU_S. 
         For certain host, why does pinging to them not work above
         a certain ping size?

Professor Ott's website:

-----------------------

CIS656 Fall 2001:

         TCP Tutorial. Gives more on RTT Estimation and Time-Outs 
         than I gave in class (still not the whole story!). For even 
         more, use Stevens & Wright Vol II. 
         Students in CIS656 by now (03/30/02) know enough about TCP 
         that they can read the tutorial.

         ISP Traffic Study. Shows how IP Traffic Traces can be used to
         build models for ``Human Source Behavior''. (That is, the
         behavior of the  HUMANS  using the computers). By now 
         (03/30/02) students in CIS656 (almost) know enough to read this.
         Read the first few pages. If you get hooked, read the rest.

Papers:

         The ``square root'' paper is important, but the math is too
         hard for most CS students. Give it a try anyhow. Don't try 
         to understand the proofs.

         The ``ECN'' paper is important. The math is still hard.
         The paper makes a start explaining what needs to be done in
         ECN. Don't try to understand the proofs.

         The ``Milcom 2001'' paper. Easy to read, puts much of the
         previous two papers in context. Recommended reading.

         The ``SRED'' paper. Also easy to read, also provides context.
         Recommended.

Other papers on the website: feel free to browse.

---------------------

Students who want to specialize in Computer Networking should know about 
WFQ (Weighted Fair Queueing). Further work on ECN (e.g. joint design 
of computer source behavior and router behavior) is likely to require 
understanding of WFQ. 

Reading on WFQ (I hope to find a better first reading):

Parekh, A.K. and Gallager, G.R. (1992) A generalized processor sharing 
approach to flow control - the single node case. Infocom 1992

Golestani, S.J. (1994) A self-clocked fair queueing scheme for broadband 
applications. Infocom 1994.


Students who want to specialize in Computer Networking should know
about Integrated Services (``IntServ'') and Differentiated Services
(``DiffServ'').

There are RFCs on these topics. Go to the IETF site and search.

Future work on ECN is likely to require some (probably not too much ?) 
understanding of DiffServ.


When designing ``router behavior'' and ``computer source behavior'' 
together, it is important not to optimize for a specific situation,
for example for a specific network topology or a specific behavior 
of the human customers. Such optimization is likely to lead to 
mechanisms that are rather bad in other situations.

Instead, we must design ``router behaviors'' and ``computer source 
behaviors'' that are quite robust over a wide range of plausible 
situations.

----

``Human Source Behavior'' vs ``Computer Source Behavior''.

The expression ``source behavior'' can mean several different things.

It may mean ``computer source behavior'': how does (for example) a TCP 
source react to an apparently lost packet? To a marked packet? etc.

It may also mean ``human source behavior'': how long does a human spend 
``thinking'' before asking for the next page or file? What kind of files
does he/she ask for?  


We can  SET  computer source behavior. We can only observe, and try to
understand,  human source behavior.

An example ``human source behavior'' is postulated in the SRED paper.


Good Luck!  Teun Ott.