Re: [apps-discuss] I-D Action: draft-ietf-appsawg-acct-uri-05.txt

[Bob Wyman <[e-mail]>]

On Tue, Jul 2, 2013 at 5:11 PM, DataPacRat <[e-mail]> wrote:

> On Tue, Jul 2, 2013 at 4:21 PM, Bob Wyman <[e-mail]> wrote:
> > On Mon, Jul 1, 2013 at 1:19 PM, Markus Lanthaler <
> [e-mail]>
> > wrote:
>
> >> I'm wondering whether it would make sense to add a feature allowing
> >> associate a date to an account. This would address problems arising from
> >> account recycling (think Yahoo). Maybe something like
> >>
> >>    acct:[e-mail]?date=20130701
> > It seems to me that if one wanted to use dates as part of identifiers,
> you
> > would probably want to learn from tag URI's as defined in RFC 4151. Tag
> > URI's handle the problem of creating identifiers that are unique across
> time
> > -- which appears to be the use-case you're looking to address.
> >
> > Thus, rather than the syntax you suggested, you'd be using something like
> > this: (i.e. use a comma, not a question-mark.)
> >
> > acct:[e-mail],2001-09-15
> >
> > The relevant ABNF bits from RFC4151 are:
> >>
> >> taggingEntity = authorityName "," date
> >> authorityName = DNSname / emailAddress
> >> date = year ["-" month ["-" day]]
>
> While being inspired by tag:, you needn't feel limited by its
> particular syntax. The date field on tag doesn't allow any more
> precision than midnight on a particular day; and it only allows for
> indicating particular moments rather than periods of time.

I think we'll find that support for "periods of time" is really not
necessary for reasonable use-cases. What the date does in this context is
allow you to say: "I am referring to the account that was named '
[e-mail]' during 2001-09-15." That should result in a single account
being identified (if we assume that account names aren't reused more than
once a day...)
If ranges were supported, someone might construct something that said: "I
am referring to any of the accounts that were identified as '[e-mail]'
between 2001-01-01 and 2013-0101." Depending on how often account names
were reused, that could refer to a large number of accounts (i.e. >1).

It is tempting, whenever dates are discussed, to want to support the full
set of ISO specified date goodies, however, I really don't think it makes
sense in this context. Do we want to encourage acct names to be reused more
than once a day? Do we want to figure out how to address returning multiple
results in protocols that use acct? (What would WebFinger do if it needed
to return multiple results?)

People will typically have an identifier that was discovered by them on
some date in the past. Using something like WebFinger, what they will be
interested in asking is: "How do I contact the person or entity who used
the account "[e-mail]" on July 2, 2013?" I assume that it would be
possible for a WebFinger user or acct: processor to respond with some data
pointing to a current email address, etc. for the user or even some
indication that they are no longer living...



> It should
> be simple enough to draw on ISO 8601's full range of time-indicators,
> to allow (while not requiring) the indication of any particular second
> (eg, 2001-09-15T12:34:56Z), or a period of time (eg, 2001-01-01/P1Y to
> indicate the entirety of that year).
>
>
> Thank you for your time,
> --
> DataPacRat
> "Then again, I could be wrong."
>