RE: general question

> I'm leary of a third-party device holding the data (can you say,
> "key-escrow?" or WellKnownPointsofFailure? - sure you can.
Well, actually, I'm thinking that I can build a server on a big Linux box
that gives me mucho better security than the puny thing I could do
on a cell phone. Can you say "PublicKeyCryptoOperationsPerSecond"?
My cell phone tells my server where I am, my server tells whomever I
authorize where I am. It has a lot more space for rules and a lot more
horsepower for crypto.

> In anycase,
> Ok. The OOD has what attributes wrt the "geo" component of geopriv?
>
> lat/log/alt
> vector/velocity
> mass/"size"
Need the other form:
planet/country/state/county/city/street/building/floor/room
Need both in the report, plus a flag that tells if one value
was translated into the other, and another field for who did the
translation.

>
> And the priv part... :)
> Classes of "who wants to know"
>
> Public data - anyone/thing can get the bits
> "Good Guys"
> "Bad Guys"
> "Gargoles" (instrumentation/measurement)
>
> (and what subclasses of these "buckets" are useful?"
> e.g. multi-juristdictional transit bwtn the requestor and
> the OOD)
Starting to get weird on me. Seems to me that you
have to have your own classes (the spec doesn't define them).
You put a requestor in a class. This might be done
on the fly, or more likely, beforehand. I like the
idea of having a way that a third party certifies that
some requestor has some characteristic you want to know,
and you can then use the certification to authorize into
a class. For example, if there were someone who would
certify GEN-U-INE pizza shops from snake oil salesmen,
that would be nice.

Anyway, a class gets a rule, and the rule says what the
requestor gets.

You have to decide what ACCURACY each class gets to each
component, where you have a value for - "no information
and let him know that", and another for "make something up
that looks reasonable, but is pure fantasy and give the
sucker that"
>
> When does the OOD release i(or authorize, in the event of third-party
> state-keepers) the data?
>
> Timeliness
> Relevence
> "granularity"
I'd express this as:
"accuracy, update rate, and minimum-change. I'm assuming by
granularity you are trying to filter out reports when a component
doesn't change by some defined amount. I'm not sure I fully
appreciate "Timeliness". I don't think it is helpfull to me to
give anyone old, but accurate location.

>
> Please fill in the missing catagories and then some kind soul can
> construct (small) matrix of useful choices.
>
> e.g. for granularity:
>
> exact match on "size"+vv
> w/in 1m of "size"+vv
> w/in 100m of "size"+vv
> w/in 10km of "size"+vv
>
> or something along these lines.
Confusion. I don't report how near I am to anything.
I can turn location around and subtractl; you tell me
where you are, I know where I am, I subtract, and I
know how far you are from me -- I don't reveal where
I am to you. Two cannot play that game together!

I do want to support tracking, which implies periodic
reports. I need to have a decision on how often I will
permit reports to be given (min period). That is
my decision. The requester can ask for a longer
period, and he can also ask for a filter on minimum
change before a report is generated. Note that
min period is important even if I don't explicitly
support tracking - I want to control how often I
will respond to a request for location, as I want to
control implicit tracking.

Brian
Received on Mon Sep 10 17:51:12 2001