I agree with Christian that there are two orthogonal issues. Comments
in line...
- Ralph
On Apr 22, 2009, at 1:19 AM 4/22/09, Christian Vogt wrote:
Folks -
It seems that folks are considering two related, yet still orthogonal
topics for inclusion in the MIF charter:
- Conflicts between configuration parameters.
- Issues with address selection.
(These two topics also span "issues with multiple network
connections",
which has been brought up as well.)
The first topic talks about a problem of the network stack: How are
conflicting parameters for host or interface configuration reconciled
with each other, or, if not reconcilable, which parameters win? This
issue may come up if configuration parameters are received from
multiple
sources, such as from any set of default routers, DHCP servers, and
manual configuration. Hosts with multiple interfaces are more
likely to
be confronted with this issue, but the issue may also come up for
hosts
with a single interface.
Yes...this topic arises whenever there are multiple sources of
"information", and those sources of information may be delivered over
a single "interface".
I think the topic is modified by the existence of multiple
"administrative domains". We could hand-wave away the topic of
conflicting information from multiple sources in a single admin domain
as "misconfiguration". Information from multiple admin domains might
be conflicting, or might be appropriate only for traffic related to
the admin domain. Roughly speaking, there is a vague description of
the related problem of per-interface and host-wide information
supplied through DHCP.
The second topic talks about a problem of applications: When
initiating
a connection, which pair of source and destination address (and
consequently which pair of interfaces) should be used? Again, this
issue may come up independently of whether a host has one or multiple
interfaces -- though the presence of multiple interfaces makes the
issue
more difficult to tackle and, since it determines which interface pair
is used, also more important.
In some circumstances, the order of selection might be reversed: pick
an interface (WiFi versus wired Ethernet) and then choose addresses.
Hm. Perhaps that would be better expressed as "order of preference";
some destination might not be reachable through one interface, even if
that interface might be preferred for other reasons.
Both topics need work; the arguments that have been put forth clearly
show this. And we have to make a decision whether to tackle either or
both of these topics in the MIF working group. Personally, I feel
that
the workload of both topics will be manageable for a single working
group. Though, if we decide to tackle both, then the working group
charter would have to explicitly list these topics as two separate
ones.
This would require some, but not much, rewriting of the charter.
In any case, the working group acronym could actually stay as is. If
the working group will tackle both topics, we could re-use the acronym
for "Multiply Interface Addressing and Configuration". If only one of
the topics will be tackled, then we could just remove from this the
words "and Configuration" or "Addressing and", respectively.
- Christian
_______________________________________________
mif mailing list
mif(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/mif
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf