Re: mech-v2-05pre

[Pekka Savola <pekkas@netcore.fi>]

On Thu, 26 Aug 2004, Alex Conta wrote:
> I suggest two small changes: remove 'the', and replace 'as' with 'of 
> the' as in the following text:
> 
>     All tunnels are assumed to be bidirectional, behaving as virtual
>     point-to-point links between nodes, using IPv4 addresses of the
>     tunnel endpoints.

Let's discuss this in another thread off-list, as it seems more of
wordsmithing than anything else.

> I suggested "automatically selected", to balance, and for symmetry to
> "manually configured". Even if you remove "automatic selection", if it 
> is not "manually configured", then it is still automatically selected, 
> so you say "manually" for "configuring", but you say nothing for 
> "automatic selection".
> 
> If you dislike "automatically selected", then just remove "manually", to
> balance the text, but to me, the text that says both is better balanced.

I could replace "manually configured" by "configured by the
administrator", if that would sound better without "automatically
selected".. otherwise I'd prefer to leave it as is.

I agree that the implementation has to select in any case (because it
says "an address"; that's because I didn't want to open the discussion
which would ensure if it said "the address" if you have multiple
addresses), but it just needs to be picked in any case, and it seems
worse to spell it out because it's suboptimal in the case when there
are multiple addresses.

> Furthermore, regarding the technical aspects, which you elaborated 
> above, I would like to walk through, to make some clarifications:
> 
> The "automatic selection of source address" relies on *route* selection:
> this is always deterministic. The selected route to destination yields 
> the outgoing interface, which is also deterministic.

Not necessarily if you run routing protocols.  Route changes somewhere
in the net could change the outgoing interface to be more optimal,
while the original interface's addresses could still work as well.
 
> So this need better clarification in the text.
> 
> Maybe changing the second phrase from:
>     This may be a problem
>     with multi-addressed, and in particular, multi-interface nodes,
>     especially when the routing is changed from a stable condition, as
>     the source address selection may be adversely affected.
> 
> to:
[...]
>     Configuring the source address is appropriate particularly in cases
>     in which automatic selection of source address is not deterministic.

This seems OK to me, though I'm concerned that just saying "not 
deterministic" might not give sufficient guidance to the readers.

Maybe it should be elaborated with another sentence like:

 Configuring the source address is appropriate particularly in cases
 in which automatic selection of source address is not deterministic;
 this is often the case, e.g., with multiple addresses or with 
 multiple interfaces when using routing protocols.

> Note that an implementation has multiple choices:
> 1. can always give a WARNING to the operator, if the "automatic 
> selection" is not deterministic, or
> 2. return with a question to the operator, which of the addresses in a 
> possible list to choose.
> 3. return the selected address to the operator, to be used for further 
> configuration at the other end node.
> 
> These could be documented in an Appendix, if you intended to have an 
> Implementation Notes section.

I'd rather avoid adding such a section.  

In this particular case, I think all the implementations either
require the admin to always specify the source address, or leave it up
to the responsibility of the admin to know when it must be specified.

-- 
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings