Bug 2120 - 802.11g networks are not compatible with 802.11b clients
802.11g networks are not compatible with 802.11b clients
Status: RESOLVED FIXED
Product: ns-3
Classification: Unclassified
Component: wifi
ns-3-dev
All All
: P5 normal
Assigned To: sebastien.deronne
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2015-05-17 06:07 UTC by sebastien.deronne
Modified: 2016-02-17 16:42 UTC (History)
3 users (show)

See Also:


Attachments
proposed documentation patch (1.23 KB, patch)
2015-09-14 10:30 UTC, Tom Henderson
Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description sebastien.deronne 2015-05-17 06:07:00 UTC
It has been observed that a 802.11b client is not supported by a 802.11g network. The issue seems to come from the association process.
Comment 1 Tom Henderson 2015-09-14 09:07:37 UTC
Is there much need to support this mixed technology use case anymore?

Related to this, there is a comment in WifiPhy::ConfigureStandard for 802.11g that short slot time is not supported.  If we do not have mixtures of 802.11g and 802.11b, should we be instead configuring a slot time of 9us by default for 802.11g?
Comment 2 sebastien.deronne 2015-09-14 09:56:23 UTC
Actually yes, in my own simulations I was fixing it to 9 microseconds. In pure g networks, it should indeed be fixed to 9us.

The mixed technology is needed, I already got some requests from people willing to consider a network with both 802.11b and 802.11g clients, and I believe it's a "nice to have".

I plan to add support for mixed b/g around the end of the year (unless someone else takes up the story), and dynamic adaptation of the slot time value to 20 microseconds if a 802.11b stations is joining the network.
This should not be too much work.
Comment 3 Tom Henderson 2015-09-14 10:30:46 UTC
Created attachment 2144 [details]
proposed documentation patch
Comment 4 Tom Henderson 2015-09-14 10:31:16 UTC
(In reply to sebastien.deronne from comment #2)
> Actually yes, in my own simulations I was fixing it to 9 microseconds. In
> pure g networks, it should indeed be fixed to 9us.
> 
> The mixed technology is needed, I already got some requests from people
> willing to consider a network with both 802.11b and 802.11g clients, and I
> believe it's a "nice to have".
> 
> I plan to add support for mixed b/g around the end of the year (unless
> someone else takes up the story), and dynamic adaptation of the slot time
> value to 20 microseconds if a 802.11b stations is joining the network.
> This should not be too much work.


Do you agree to the attached documentation patch, in the meantime?
Comment 5 sebastien.deronne 2015-09-14 10:32:44 UTC
Yes it's fine.
Comment 6 sebastien.deronne 2015-11-24 14:43:49 UTC
A review request have been started for changes related to this bug:
https://codereview.appspot.com/273510043

Note that documentation will be updated once the changes are being delivered.
Comment 7 sebastien.deronne 2015-12-15 15:15:39 UTC
Note that I plan to update documentation accordingly in the middle of next week, so I would be glad if reviews could be handled this week.
Comment 8 sebastien.deronne 2016-01-31 14:49:28 UTC
I have not yet handled code review comments, because I cannot reproduce scenarios where protection show any improvement...

Even though I slightly improved the implementation since latest patch set, I do not observe any bugs from traces, so I am really wondering in which case protection mechanism really enhance something!

I am still searching from existing studies in the literature, but I have not yet found something significant, maybe this has never really been studied? I am a bit surprised because this is already a quite old feature. 

Any help/tip is welcome :-)
Comment 9 sebastien.deronne 2016-02-07 17:20:35 UTC
(In reply to sebastien.deronne from comment #8)
> I have not yet handled code review comments, because I cannot reproduce
> scenarios where protection show any improvement...
> 
> Even though I slightly improved the implementation since latest patch set, I
> do not observe any bugs from traces, so I am really wondering in which case
> protection mechanism really enhance something!
> 
> I am still searching from existing studies in the literature, but I have not
> yet found something significant, maybe this has never really been studied? I
> am a bit surprised because this is already a quite old feature. 
> 
> Any help/tip is welcome :-)

All references I could find confirm the results: protection mechanism is needed to ensure a correct NAV setting for 802.11b in case of an ERP OFDM transmission, but at a cost of a throughput degradation.

I will shortly post a new patch set.
Comment 10 sebastien.deronne 2016-02-08 17:32:58 UTC
New patch set available
Comment 11 sebastien.deronne 2016-02-14 10:43:09 UTC
Another patch set available.
If no remaining comments, I will deliver those changes by February 18th.
Comment 12 Tommaso Pecorella 2016-02-14 10:50:30 UTC
(In reply to sebastien.deronne from comment #11)
> Another patch set available.
> If no remaining comments, I will deliver those changes by February 18th.

Deadpool's movie day !
Comment 13 sebastien.deronne 2016-02-17 16:42:06 UTC
changeset 11881:8707c44ecc30