[ClusterLabs Developers] fencing-topology change in 1.1.14 (especially for pcs/crm authors)

[Ken Gaillot]

On 09/08/2015 01:51 PM, Jan Pokorný wrote:
> On 08/09/15 10:27 -0500, Ken Gaillot wrote:
>> * Don't confuse NAME:VALUE with an IPv6 address, which must have
>> more than one ':'
> 
> conceptually not enough to reliably distinguish IPv6 address 
> "abcd:ef::1" and restriction on "abcd" node attribute being
> "ef::1" unless we claim the attribute cannot contain more than a
> single colon if not in prefix, none otherwise (or variations on
> this).

Good point. The current Pacemaker logic will treat any target with no
equals and multiple colons as an IPv6 address, so tools can follow
that same rule, or be more intelligent as you suggested below, as long
as what they end up giving Pacemaker follows that rule.

> In fact I can see that colon is completely excluded in VALUE from 
> NAME:VALUE notation, hence when configuration tools observe 
> TARGET_SPEC akin to this NAME:VALUE, where VALUE contains 1+ colon 
> and whole TARGET_SPEC is not lexically correct IPv6, they could
> either: - turn NAME:VALUE into NAME=VALUE silently - complain about
> invalid IPv6 + "did you mean NAME=VALUE instead?"
> 
> Of course, this assumes no extra syntax is invented with its own 
> set of restrictions, but that should watch out for the same under 
> the hood.
> 
> Just and idea what to be careful about here.