Re: UDP or TCP?

----------
] From: Jeffrey Mogul  <mogul@pa.dec.com>
] To: Paul Leach
] Cc:  <netmail!http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com>;  <mogul@pa.dec.com>
] Subject: Re: UDP or TCP?
] Date: Friday, August 11, 1995 11:44AM
]
]     ] But simply going to UDP is not the answer, because UDP (without
]     ] some extra effort) does not offer any congestion-control mechanisms.
]     ] Sure, you can layer them on top of UDP, but I suspect that you would
]     ] quickly get back to something very similar to TCP.
]
]     All RPCs I know of do the extra flow control thing. For RPC, it is
]     very simple in the basic request/response case, as request/response
]     is naturally flow controlled -- this is why you don't quickly get
]     back to TCP for simple cases. When the in or out arguments are
]     large, then you do indeed have essentially the same problem as
]     TCP.
]
] Flow control is not congestion control.

I mis-wrote.  I meant to say congestion control. If a sender sends one 
packet, and waits for a response before sending another, then 
congestion automatically slows down the rate at which the sender transmits.

]
]     ] Paul, what does DCE RPC do for congestion avoidance and control
]     ] in WANs, when it uses UDP?
]
]     The protocol is designed to use slow start, sliding window, ack
]     maps, and serial numbered packets.
]
] Slow start (depending on exactly how it is done) will help with
] congestion.  It may not be a full solution.

Let me restate: the combination of the above items is everything TCP 
does, and more. ACk maps reduce retranmissions, and serial numbered 
packet reduce retransmission and make round trip time estimation more accurate.

]
]     I don't know how much of the protocol capacity is
]     exploited in any given implementation -- MS, OSF, and OSF members'
]     "enhanced" versions.
]
] If congestion control is not a mandatory part of the implementation,
] present in all cases, it would be unwise to deploy this as the basis
] for the Web.

I agree. Before using this protocol, the rules for an implementation 
that would be used for the Web should be tightened a lot.  An almost 
literal copying of the relevant portions of the TCP spec into its spec 
would do the trick. And the reference implementation should of course conform.

Paul

Received on Friday, 11 August 1995 14:29:12 UTC