[e2e] TCP Option Negotiation
Ralph Droms
rdroms at cisco.com
Thu May 17 09:53:50 PDT 2001
I've had discussions about enforcing a quiet time in DHCP - a minimum delay
time before reassigning an address to a new host.
We've never heard a sufficiently strong argument to warrant adding the
requirement to the DHCP spec. I don't know of any servers that implement a
quiet time.
- Ralph
At 12:25 PM 5/17/2001 -0400, John Wroclawski wrote:
>At 11:57 AM -0400 5/17/01, Hari Balakrishnan wrote:
>>> Alex,
>>>
>>> This seems to be another manifestation of the standard problem of old
>>> duplicate packet. Your scenario is a violation of TCP's "quiet time"
>>> requirement upon host crash and restart (it's the same host if it has
>>> the same IP address). Quiet time is a vital part of TCP's machinery to
>>> protect against old duplicates.
>>
>>Bob,
>>
>>Not quite.
>>
>>Unfortunately the statement: "it's the same host if it has the same IP
>>address"
>>is increasingly untrue because of dynamic IP address assignment (e.g., via
>>DHCP). This may well be a theoretical problem, but I've observed (in my
>>home,
>>from my FreeBSD DHCP server), turning off a laptop and turning another
>>one on,
>>and having the latter receive the former's IP!
>>
>>Hari
>
>Hari,
>
>It might be arguable that RFC793 actually covers this. Crash/restart is
>used as the motivating example, but the words are more general - "in the
>absence of knowledge about the sequence numbers used on a particular
>connection, the TCP specification recommends that the source delay for MSL
>seconds before emitting segments on that connection, to allow time for
>segments from an earlier connection incarnation to drain from the system".
>
>An implementation that truly followed this recommendation would enforce a
>quiet time after any assignment of a dynamic address on the grounds that
>it had no idea where that address had been before, and thus no knowledge
>about previous SN's..
>
>John
>
More information about the end2end-interest
mailing list