[PATCH xserver] XDMCP: For IPv6 add IPv6 link local addresses to the end of the list

[Reinhard Max]

On Tue, 21 Jun 2016 at 17:56, Stefan Dirsch wrote:

> On Tue, Jun 21, 2016 at 11:46:05AM -0400, Adam Jackson wrote:
>> On Fri, 2016-04-29 at 11:22 +0200, Stefan Dirsch wrote:
>>> On Fri, Apr 29, 2016 at 09:40:05AM +0200, Julien Cristau wrote:
>>>> On Thu, Apr 28, 2016 at 15:21:15 +0200, Stefan Dirsch wrote:
>>>>
>>>>> From: Reinhard Max <max at suse.de>
>>>>>
>>>>> For IPv6 add a link local addresses to the end of the list 
>>>>> passed to the XDMCP servers. Reason: for link local addresses 
>>>>> the XDMCP server would need to either know the interface thru a 
>>>>> scope identifier or try all available interfaces. If they don't 
>>>>> this address will fail in which case the XDMCP server could 
>>>>> still try the other addresses passed - however some only try the 
>>>>> first address and then give up. Even if this seems to be the 
>>>>> wrong place to fix this it seems to be easier than fixing all 
>>>>> display servers.
>>>>
>>>> Commit message says "add to the end of the list", code says 
>>>> "ignore", which is it?
>>>
>>> Good question. Maybe Reinhard still remembers. Meanwhile I found 
>>> slightly different descriptions for the patch. Seems the patch 
>>> changed in between. Sigh.
>>
>> Ugh. I found this patch in the patchwork backlog and pushed it 
>> before checking for this thread; reverted.
>>
>> The logic in the patch does seem useful; is there a final version 
>> you're shipping, and does Max remember the details yet?
>
> It is the final version we're shipping - since years. Reinhard is in 
> Cc. He may reply in case he still remembers the details. Thanks for 
> asking!

Sorry for not replying earlier, this had somehow fallen off my radar.

It has been too long for me to remember exactly how the discrepancy 
between the commit message and the code came to be. I guess that while 
working on the patch I changed my mind and suppressed the link local 
addresses instead of pushing them to the end of the list, but then 
forgot to update the patch header, which can easily happen when one 
modifies existing patches with quilt.

What I do remember is that back in the days none of the display 
managers was able to handle link local addresses properly and hence 
suppressing them seemed to be the way to go.

If the situation has improved since (I haven't seen any DM code for 
years) and some DMs can now deal with LL addresses, it might make 
sense to do what the commit message says and push them to the end of 
the list. They should just not be on the top of the list to cater for 
DMs that only try the first address and can't deal with LL.

I hope this helps somewhat.

cu
 	Reinhard