On Thu, Jul 17, 2003 at 02:37:19PM -0400, Greg Troxel wrote: > > I have been able to depend on Prism firmware versions 1.4.9 and 1.5.6 > > to do WECA-compliant IBSS merges. > > Mine is 1.4.9. Have you seen the 'changing middle two octects" > behavior? Is this a symptom of a 1-member IBSS?
I have seen that. I think that it is a symptom of a 1-member IBSS. > > I have no idea which versions of Lucent firmware do a proper merge. > > I had 6.16.1 and 7.28.1. Both seems to have no issues, but I haven't > tried n of each card, etc. BTW, to verify IBSS merges, it is desirable to have two IBSSs out of radio range of each other, each IBSS consisting of two or more stations in a "steady state" (i.e., constant BSSID, not scanning). > > Note that under the WECA rules, two IBSSs which have formed with the > > same SSID in the same airspace but on different channels do not have > > to merge. Also, a station with a choice of IBSSs with the right SSID > > will usually join the IBSS which sends the loudest beacon or probe > > response instead of the oldest IBSS (the IBSS with greatest TSF). > > But then the two IBSSs will be in mutual range, and the merging rules > should require a cascade to the older one. I should have been more clear. Change the above to "a station with a choice of IBSSs with the right SSID _on different channels_." As I mentioned above, WECA only requires a merge of IBSSs on the same channel. > It sounds like after the merge, if there is a partition, there will > then be two copies of the same IBSS, but each with a separate master > (same BSSID). There could then be TSF drift, and there should be a > new master elected once the partition is healed. There is no "master" station. The whole notion of "create IBSS"-mode and "IBSS master" is a Lucent- or Prism-ism. > > I doubt that the ADMtek chip does IBSS merges by itself. In my ADMtek > > driver, I have put some rudimentary logic into atw_recv_beacon for > > detecting and joining the oldest IBSS on the channel which has the > > correct SSID. I do not account for Rx and other latencies, yet. > > It sounds like this is a host/driver function, which means one could > adjust it to match the spec if it doesn't. Right. Dave -- David Young OJC Technologies [EMAIL PROTECTED] Urbana, IL * (217) 278-3933 -- general wireless list, a bawug thing <http://www.bawug.org/> [un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless
