Re: [homenet] RFC 7788-bis (and also draft-cheshire-homenet-dot-home-03)

[Andrew Sullivan <[e-mail]>]

Dear colleagues,

On Fri, Jun 17, 2016 at 09:37:06AM +0100, Ray Bellis wrote:

> Whilst there may be "undermined" ways it's being used, it's clear that
> most of the ways it's used are just because some vendors and sites
> decided to use that for their default *site local* domain which makes it
> completely consistent with what we need.
> 
> I therefore completely disagree on point #1 - officially allocating
> .home for this purpose and having it "sunk" by default on internet
> facing recursive resolvers would IMHO actually *help* with the traffic
> hitting the root and reduce leakage of it.

The argument above, which is rehearsed to some extent in
draft-cheshire-homenet-dot-home-03 (although there IMO more subtly),
is interesting to me, in that it appears to start with the exact same
premises I do and reach the exact opposite conclusion.  I believe this
is because of some unstated premises, and so I'm going to attempt to
lay out the premises as I understand them as completely as I can.

To do this, I'm going to draw some inferences about what Ray was
arguing and about what is in draft-cheshire-homenet-dot-home-03.  I
hope the authors indulge me, and I hope you, Gentle Reader, do not
mistake my inferences as speaking correctly either for Ray or for
Stuart.  Since they're both here, they can correct what I get wrong;
but I think trying to lay out this different story might help us.

I think we all agree that the label home, in the top-most position of
a domain name (but maybe not a name in the DNS), is in use by some
people.  I think we all agree that at least some uses of that name are
somehow related to "stuff in my house behind my home-router-like
thing".  And I think we all agree that, whatever basis for that use
is, it either is not or ought not to be related to any actual
delegation of the name in the DNS.

With the above premises, I conclude that home is by definition not
suitable for our purposes.  I conclude that from these additional
premises:

    • that, given the detectable pollution of the namespace at and
      beneath home, there is a significant population already using
      the name for some purposes, we know not what;

    • that if we want an identifier to be some sort of protocol switch
      by which we tell software to do something novel, we need an
      identifier that has at least a modest chance of not running into
      widely-deployed use for some purpose not defined to be
      consistent with the protocol switch;

    • that it is at least fantastically difficult to suss out all the
      strange things people are already doing with "in the wild"
      undelegated names in the DNS, even if we make the dubious
      assumption that there is something like a rigorous design behind
      those doings;

    • that strings that could be used as protocol switches are
      fundamentally machine-directed rather than human-directed, and
      therefore have a certain arbitrariness about them.

With the same premises, I think the opposite argument is that home is
entirely good for our purposes, because of the following additional
premises:

    • that we have (or we can get, which is what
      draft-cheshire-homenet-dot-home-03 is asking for) a pretty clear
      idea that all the uses of home are already more or less what
      we're trying to do;

    • that picking the same string is very unlikely to break any of
      the existing behaviour;

    • that a meaningful string to a human user is of high importance
      here;

    • that a primary (or even important secondary) motivation for the
      allocation would be to capture traffic that should never have
      been destined for the root in the first place;

    • that with adequate documentation, a possibly-conflicting use of
      home would not have negative effects.




I think the effort to document what people are actually doing with
home is laudable, and I hope it succeeds in producing a more-or-less
complete account of the use of that name.  But I do not see how we can
get from "documenting these uses is good" to "having documented it,
you can then use the name that way."  The kind of exhaustive survey
that would be needed to show the real uses of home would cost far more
in time, effort, and money than the convenience of the string
presents.  Moreover, it's not even clear that this would be the
"right" string.  For lots of people on Earth don't use Latin writing,
never mind English words.

I hope this explains why I think proceeding with home is problematic.

Andrew (speaking only for myself).

-- 
Andrew Sullivan
[e-mail]