[LINK] FTTP soon normal

[Roger Clarke]

At 23:32 +1000 28/4/14, Hamish Moffatt wrote:
>There's no CDMA on a switched network...

C'mon Hamish.  You know he meant CSMA/CD.    (:-)}

Lots of us have had that slip of the tongue.

____________________________________________


>On 28/04/14 19:29, Frank O'Connor wrote:
>>  Sorry,
>>
>>  I meant CDMA ... collision detection. Error correction, as Hamish said
>>  ... is a level 3 Feature.
>>
>>  Again ... just my 2 cents worth ...
>>  ---
>>  On 28 Apr 2014, at 6:39 pm, Hamish Moffatt <hamish at cloud.net.au
>>  <mailto:hamish at cloud.net.au>> wrote:
>>
>>>  On 28/04/14 18:22, Frank O'Connor wrote:
>>>>  Well, yeah ... but:
>>>>
>>>>  1. ANY form of networking causes 'slow-downs' simply by its very
>>>>  nature, irrespective of what the data interface is capable of.
>>>>
>>>>  1Gbs hard wired Ethernet? Sure ... if you only have 2 devices
>>>>  connected, are running a single networked application ... and even
>>>>  then all you'll get is 300-500Mbs max due to error correction (huge
>>>>  overhead in Ethernet which increases logarithmically as nodes
>>>>  activate), data scheduling problems and lots of negotiations
>>>>  (e.g.ACK/NACKS, non-data packets ... ICMP for example, and other
>>>>  high level protocols inherent in TCP/IP) between the devices.
>>>>
>>>>  It doesn't much matter what network architecture you use ... the
>>>>  overheads persist (as they were designed to do by the network
>>>>  protocol inventors) and slow traffic way below the optimum. With
>>>>  networks its important that little numbers like error detection and
>>>>  recovery work ... especially in non-tolerant applications and devices.
>>>
>>>  I think you're getting your layers pretty mixed up here.
>>>  1000base-T/802.3ab (Gigabit Ethernet) has no error correction (it has
>>>  error detection), and given that's it's almost always switched won't
>>>  have problems scaling as you add more devices and applications unless
>>>  your switch is completely hopeless. Of course it has overheads that mean
>>>  you won't actually get 1000Mbit/sec of user data (HTTP or whatever) but
>>>  the performance is pretty predictable and quite close to the theoretical
>>>  with modern computers.
>>>
>>>  TCP/IP adds overheads to get its work done but there's no interference
>>>  between nodes and applications there either.
>>>
>>>>
>>>>  2. Bottom line: WiFi is no more or less efficient than hard wired
>>>>  network protocols. Indeed, low level WiFi protocols are typically
>>>>  Ethernet protocols ... and hence subject to the SAME efficiency and
>>>>  effectiveness limitations as the wired protocols they emulate. The
>>>>  difference is that with WiFi you can overlay channels more easily
>>>>  than you can on an Ethernet connection ... which doesn't handle
>>>>  packet crowding very well at all.
>>>
>>>  WiFi of course is working on a shared channel, while switched ethernet
>>>  effectively has a separate channel for each connection.
>>>
>>>
>>>
>>>  Hamish
>>>  _______________________________________________
>>>  Link mailing list
>>>  Link at mailman.anu.edu.au <mailto:Link at mailman.anu.edu.au>
>>>  http://mailman.anu.edu.au/mailman/listinfo/link
>>
>
>_______________________________________________
>Link mailing list
>Link at mailman.anu.edu.au
>http://mailman.anu.edu.au/mailman/listinfo/link

-- 
Roger Clarke                                 http://www.rogerclarke.com/

Xamax Consultancy Pty Ltd      78 Sidaway St, Chapman ACT 2611 AUSTRALIA
Tel: +61 2 6288 6916                        http://about.me/roger.clarke
mailto:Roger.Clarke at xamax.com.au                http://www.xamax.com.au/

Visiting Professor in the Faculty of Law            University of N.S.W.
Visiting Professor in Computer Science    Australian National University