User-Agent: Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.8b2) Gecko/20050626 MultiZilla/126.96.36.199n Build Identifier: Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.8b2) Gecko/20050626 MultiZilla/188.8.131.52n While I was able to do simple DNS lookups from the command line, using nslookup and dig connected to the default Vodafone (that's who supplies the hotspot here at the Mercure Newa Dresden, and it runs on a Cisco backbone which is apparently IPv6 capable) server and received fairly quick results, attempting to browse pages - even within the captive portal - proved problematic. POP3 and SMTP were all but impossible. In desperation, I changed my resolv2 to point to a couple root servers on the net, and restarted Mozilla. Immediately, I was able to browse, and download and send outgoing mail. I then changed network.dns.disableIPv6 to true in prefs.js, restarted the browser, and finally an interface (re)configure from XWLAN (released and renewed the DHCP params), and I was right on-line again - using the default Vodafone DNS address. Additional googling (once online again) turned up numerous tuning tips on the net (mostly related to Linux platforms) advising this very tweak. A search of bugzilla for "network.dns.disableIPv6" turned up no less than 200 hits (I concede that not all of these were specifically related to this setting, but a good many dealt with slow DNS responses and timeouts). I suggest defaulting this setting to true, at least until the kinks have been ironed out of the logic, or, if this is indeed related to OS/2's lack of IPv6 functionality, then leaving this as the default on OS/2 until such time as we have a v6-capable IP stack on OS/2. Reproducible: Always Steps to Reproduce: 1. Connect to a DHCP server which assigns the primary DNS address of an IPv6 host or set the address manually in resolv2. 2. Attempt to open a web page which has not been cached. 3. Edit prefs.js to set network.dns.disableIPv6 to true. 4. Close and restart Mozilla. 5. Repeat step 2. Actual Results: After step 2, blank page, "looking up host [...]" in status bar, or simply timeout attempting to contact host messages appear in window until network.dns.disableIPv6 is set to true in prefs.js. Afterwards, hosts resolv and pages open as they should. Expected Results: Hosts should resolve and pages should open without undue delay (i.e., DNS lookups in Mozilla should take no longer than lookups via dig or nslookup). Pages should load as rapidly as possible upon resolving an address. This also affects POP3 and SMTP, and I would expect IMAP as well as other browser-type protocols (FTP, etc.).
those test results seem to indicate that the DNS server you used didn't handle AAAA queries well, not that OS/2 has a general problem with it...
(In reply to comment #1) True, however, upon rebooting to W2K Advanced Server SP4, I had full functionality - with no prefs.js modification - under Moz, FF, and Tbird (all versions contemporary with what I'm running under OS/2). This leads me to believe that while this may indeed be a server-side problem, Windows (at least) does not seem to be affected.
that version of windows may not have IPv6 support (I am not sure if it comes with the OS or if it needs extra installation)
(In reply to comment #3) The installation I have of W2K Adv Server on my ThinkPad does not have IPv6 enabled. So, essentially, the stack in that is no more capable than the one under OS/2 (okay, so it's Windows and it's terribly unstable in comparison to OS/2's IP stack, but for these purposes, they're both IPv4-capable ONLY). All I'm saying is that for OS/2, this pref should be defaulted to true, as it took me two days of troubleshooting - at 30 Euros per day for Wi-Fi access in the hotel - to figure out why I couldn't get anywhere under OS/2 while I could under Windows. Perhaps when the logic is working better (i.e., more tolerant of broken servers), we can default this to false again. Lewis
oh, OS/2 does not support ipv6?
(In reply to comment #5) > oh, OS/2 does not support ipv6? Nope. The current IP stack is based on an older AIX stack which did not support IPv6. AFAIK, there is no plan to support IPv6 on OS/2, either (but we can hope...). Lewis
What a guy! Thanks, Mike. ;-) I suppose we should wait to change the status of this until we actually build (test) with the code you've patched? (Sorry for the newbie question...I've never gotten this far with one of my submissions.) Lewis
Not really any testing required - we know it will work :)
Comment on attachment 189003 [details] [diff] [review] Disable ipv6 for OS/2 looks like this patch contains some unrelated changes to font prefs...
interesting. My editor horked that up...
Created attachment 192174 [details] [diff] [review] Disable ipv6 for OS/2 (clean patch) As a reminder to Mike here is the a new, clean patch without the font changes. I think this should also go into the 1.7 and aviary branches? Not sure who reviews such a thing.
Dumb question (please pardon the simplicity of this) which may be rhetorical: Upon updating Mozilla, prefs.js does not get updated with the changed information, correct? IOW, if a user has Mozilla 1.7.3 installed and later upgrades to 1.7.12, and assuming this change has been checked in, the user's prefs.js do not change. Therefor, a manual change is still required unless one creates an entirely new profile (in which case, the new profile defaults will be as provided in the distro). I came across this problem once again, when my brother was using a Wayport hotspot. Lewis
Luckly, it isn't entirely like that. When Mozilla saves prefs.js, it omits any preferences that have their (current) default value . Thus, if the defaults change, the use will get the new defaults just fine, so long as they haven't changed it themselves to something else.  There are a few issues with the "omit default values" idea, documented in other bugs, but I'll spare this bug from them. :)
(In reply to comment #14) > Luckly, it isn't entirely like that. When Mozilla saves prefs.js, it omits any > preferences that have their (current) default value . Thus, if the defaults > change, the use will get the new defaults just fine, so long as they haven't > changed it themselves to something else. Ah, thank you, James. This indeed makes sense (particularly in this case, where the only change to the default would be to correct the condition in the first place). Good information. Of course, we may need to deal with the reverse situation at some point in the future, is we ever get an IPv6-aware stock on OS/2. Lewis
So should that patch be checked in? Esp. for 1.8?
Comment on attachment 192174 [details] [diff] [review] Disable ipv6 for OS/2 (clean patch) r=mkaply Requesting 1.8 approval - OS/2 only
Fix checked in. Note this was not in ff rc1.