RE: case sensitivity of XRI authorities

["Drummond Reed" <drummond.reed@onename.com>]

[Note: I moved this discussion to the Editor's list for archiving. I
also copied in Dave's second response re Unicode case rules so it would
all be in one place. =DSR]

Okay, so Dave's argument, as I understand it now, is that:

A) We should specify full Unicode 4.0 caseless matching for the XRI
authority segment because this should be true of ALL XRI authority
identifiers, this being the global portion of an XRI that uses an XRI
authority.

b) We should not specify any case insensitivity rules for the rest of an
XRI path because that should be up to each authority.

To further summarize, this means that we are proposing that the 1.0
specification will say that there are only 3 places in which
case-insensitivity applies to ALL XRIs:

1) Percent-encoding.

2) URI authority segments using the rules specified in RFC 2396.

3) XRI authority segments as specified in Unicode Standard Annex #21
(http://www.unicode.org/reports/tr21/tr21-5.html)

Does everyone agree with this proposal?

=Drummond 


***DAVE'S ADDITIONAL RESPONSE***

Incidentally, the new Unicode 4.0 spec does describe caseless matching.
The
text was incorporated from
http://www.unicode.org/reports/tr21/tr21-5.html.
There are two or three corner cases that are problematic, but the vast
majority of situations are covered.


-----Original Message-----
From: Dave McAlpin [mailto:dave.mcalpin@epokinc.com]
Sent: Thursday, October 23, 2003 8:16 AM
To: Drummond Reed; Wachob, Gabe; Lindelsee, Mike; Nat Sakimura (E-mail);
Peter C Davis (E-mail)
Subject: RE: case sensitivity of XRI authorities

Right, but that only addresses my problem if the global namespaces are
case-insensitve, which puts us right back where we started.

Imagine DNS names were case sensitive. You visit a page with a link to
http://www.Microsoft.com. You'd like to visit Microsoft's website so you
click the link. Unfortunately, www.Microsoft.com is a malicious site (as
was
the site that contained the link). You really wanted to go to
http://www.microsoft.com but were tricked into going to a rogue site
because
you, as a user, didn't know the proper case of the target site. That
would
be bad. It seems to me that case sensitive comparisons in the
XRI-authority
segment are bad for exactly the same reason.

With that said, I AM sensitive to the concern about internationalization
and
the message it sends to treat English language characters differently.
Let
me ask the question this way - if there was a good algorithm for doing
case
insensitive comparisons across the entire Unicode range, would you guys
support case-insensitivity in the XRI-authority?

Dave

> -----Original Message-----
> From: Drummond Reed [mailto:drummond.reed@onename.com]
> Sent: Wednesday, October 22, 2003 9:42 PM
> To: Dave McAlpin; Wachob, Gabe; Lindelsee, Mike; Nat Sakimura
(E-mail);
> Peter C Davis (E-mail)
> Subject: RE: case sensitivity of XRI authorities
>
>
> Dave,
>
> I find your argument compelling too. I agree that the authority
> component is a special case, that could be treated differently that
the
> rest of the XRI.
>
> The biggest problem I have with saying XRI authorities should be
> case-insensitive is that XRI authorities, unlike URI authorities,
allow
> the full IRI character range (for which I think we all agree we cannot
> define case insensitive rules, unless we can adopt something specified
> by Unicode).
>
> If that's the case, then singling out the ASCII range for
> case-insensitivity in XRI authorities seems, well, a little
insensitive
> (pardon the pun ;-)
>
> An alternative is to say that BECAUSE XRI authorities are
> internationalized, case insensitivity is not specified at the level of
> the XRI spec, but can be adopted by any particular XRI authority.
>
> How do folks feel about that?
>
> =Drummond
>
> -----Original Message-----
> From: Dave McAlpin [mailto:dave.mcalpin@epokinc.com]
> Sent: Wednesday, October 22, 2003 6:08 PM
> To: Drummond Reed; 'Wachob, Gabe'; 'Lindelsee, Mike'
> Subject: RE: case sensitivity of XRI authorities
>
> You're right that this wasn't my original position. I'm willing to be
> outvoted, but consider the following.
>
> 1) The authority segment is special, primarily because we define the
> resolution mechanics for it. I don't think it's inconsistent to define
> equivalence rules for it and be silent about non-authority segments.
>
> 2) In fact, because we define resolution, I think we're obliged to
give
> guidance about case sensitivity in the authority component, one way or
> the
> other.
>
> 3) The authority component is the piece a user or programmer is most
> likely
> to guess. That's certainly true for domain names and I'll bet it's
> equally
> true for xri authorities.
>
> 4) If the authority component is case insensitive, guessing is
> dangerous.
> For example, xri:@Nordstrom and xri:@nordstrom will resolve
differently
> (possibly to the same network endpoint, but they wouldn't be
equivalent
> by
> definition). That means that Nordstrom either needs to register all
> possible
> case distinctions a user might reasonably guess or else be subject to
> spoofers. I think both options are bad.
>
> My preference is
>
> A) Percent-encoding is case INSENSITIVE.
> B) Both URI-authorities and XRI-authorities are case INSENSITIVE.
> C) Everything else is defined by the naming authority that controls
the
> segment or subsegment in question (i.e. by its immediate parent).
>
> Dave
>
>
>
> -----Original Message-----
> From: Drummond Reed [mailto:drummond.reed@onename.com]
> Sent: Wednesday, October 22, 2003 12:58 PM
> To: Wachob, Gabe; Lindelsee, Mike; Dave McAlpin
> Subject: RE: case sensitivity of XRI authorities
>
> Ah, good, at least I think we're all on the same page now. Mike's
> argument about case sensitivity convinced me that C was the best path
> for XRI authorities, even though it's different than URI authorities.
My
> reasoning is that by not requiring in the spec for XRI authorities, we
> keep the option open for specific XRI authorities to "narrow" their
> namespace to case sensitive, but that the spec itself will be broad
> enough to allow for case-insensitive authorities.
>
> So, Dave, it seems that Gabe, Mike, and I would all prefer to adopt C.
> What's your stance?
>
> =Drummond
>
> -----Original Message-----
> From: Wachob, Gabe [mailto:gwachob@visa.com]
> Sent: Wednesday, October 22, 2003 12:33 PM
> To: Drummond Reed; Wachob, Gabe; Lindelsee, Mike; Dave McAlpin
> Subject: RE: case sensitivity of XRI authorities
>
> We are out of sync!
>
> C) is NOT agreed to. I like the position stated in C, but I think it
the
> opposite of what we've agreed to.
>
> I gather that Mike L would follow whatever decision we made, so I
wonder
> what Dave's position is on case-sensitivity of XRI-authorities. I'd
like
> to adopt the position in C)
>
>         -Gabe
>
> > -----Original Message-----
> > From: Drummond Reed [mailto:drummond.reed@onename.com]
> > Sent: Wednesday, October 22, 2003 12:29 PM
> > To: Wachob, Gabe; Lindelsee, Mike; Dave McAlpin
> > Subject: RE: case sensitivity of XRI authorities
> >
> >
> > First, my bad. My message said "So the conclusion is
case-sensitivity
> > only for percent-encoding and URI-authorities" and what I
> > meant was case
> > INsensitivity for these. Sorry, just not enough coffee this
> > morning ;-)
> >
> > Second, I'm still unclear from Gabe's message below. So let me put
it
> > this way. I believe what we agreed on was that:
> >
> > A) Percent-encoding is case INSENSITIVE.
> >
> > B) URI-authorities are case INSENSITIVE.
> >
> > C) Everything else, including XRI-authorities, are case
> > SENSITIVE. (If a
> > particular XRI authority decides to apply case INSENSITIVITY to a
> > namespace under their control, that's their perogative, but the spec
> > will not require that.)
> >
> > Does everyone agree?
> >
> > =Drummond
> >
> > -----Original Message-----
> > From: Wachob, Gabe [mailto:gwachob@visa.com]
> > Sent: Wednesday, October 22, 2003 12:03 PM
> > To: Drummond Reed; Lindelsee, Mike; Wachob, Gabe; Dave McAlpin
> > Subject: RE: case sensitivity of XRI authorities
> >
> > Wait guys!
> >
> > We're getting a little confused here.
> >
> > First of all, lets be clear about case-sentivity vs. INsensitivity.
> >
> > 1) I think we all agree that domain names (URI authorities) are case
> > INSENSITIVE and there is nothing broken about this.
> >
> > 2) The issue is about non-domain names (XRI authorities) - for
example
> > xri:+foo and xri:+Foo. Currently they are specified as case
> > INSENSITIVE
> > (meaning that US-ASCII characters as case-insensitive, but all other
> > characters, including, for example, Ñ and ñ are treated
> > differently - ie
> > case SENSITIVE). I'm not convinced this is required for URI
alignment
> > and in fact, I think its a bad idea given the Ñ/ñ situation I
> > mentioned
> > in a previous email.
> >
> > 3) And %-escaped characters are equivalent whether or not the
> > hex digits
> > are upper or lower case (ie %b0 is the same as %B0).
> >
> > Is this what everybody is agreeing to?
> >
> >         -Gabe
> >
> > > -----Original Message-----
> > > From: Drummond Reed [mailto:drummond.reed@onename.com]
> > > Sent: Wednesday, October 22, 2003 11:48 AM
> > > To: Lindelsee, Mike; Wachob, Gabe; Dave McAlpin
> > > Subject: RE: case sensitivity of XRI authorities
> > >
> > >
> > > Thanks, Mike. I recalled that your reasoning convinced me
> > and it still
> > > does. So the conclusion is case-sensitivity only for
> > percent-encoding
> > > and URI-authorities. XRI authorities can then of course
> > decide on the
> > > rules for their individual namespaces, but the spec will say
nothing
> > > about this.
> > >
> > > Done.
> > >
> > > =Drummond
> > >
> > > -----Original Message-----
> > > From: Lindelsee, Mike [mailto:mlindels@visa.com]
> > > Sent: Wednesday, October 22, 2003 10:30 AM
> > > To: Drummond Reed; Wachob, Gabe; Dave McAlpin; Lindelsee, Mike
> > > Subject: RE: case sensitivity of XRI authorities
> > >
> > > I was just reviewing that thread and Dave remembered the
> > > result of that
> > > discussion correctly.  We left it that percent encoding and
> > > URI-authorities would be case-insensitive.  I'm still not
> > > convinced that
> > > URI-authorities need always be case-insensitive (except in the
case
> > > where we are being backwards compatible with DNS names), but
> > > am happy to
> > > accept the wisdom of the group on this. ;)
> > >
> > > My reasoning, btw, was (imho) not particularly strong, but
> > still makes
> > > sense to me.  That is, unicode issues, case-sensitivity in many
> > > filesystems, and case-sensitivity in account/resource naming
> > > all lead me
> > > to feel that it would be more flexible and forward-looking to
> > > not limit
> > > ourselves by declaring all of an XRI (or even just the
> > Authority part)
> > > case-insensitive.
> > >
> > > Mike
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Drummond Reed [mailto:drummond.reed@onename.com]
> > > > Sent: Tuesday, October 21, 2003 6:48 PM
> > > > To: Wachob, Gabe; Dave McAlpin; Lindelsee, Mike
> > > > Subject: RE: case sensitivity of XRI authorities
> > > >
> > > >
> > > > I know I'm late on this but I though Mike persuaded us all that
> > > > everything should be case sensitive. Mike, what was your
> > > argument, and
> > > > where did we end up after you made that argument?
> > > >
> > > > =Drummond
> > > >
> > > > -----Original Message-----
> > > > From: Wachob, Gabe [mailto:gwachob@visa.com]
> > > > Sent: Tuesday, October 21, 2003 1:04 PM
> > > > To: 'Dave McAlpin'; Wachob, Gabe; Drummond Reed; Lindelsee, Mike
> > > > Subject: RE: case sensitivity of XRI authorities
> > > >
> > > > I haven't changed the text as it is now, lets see if we get
> > > > comments on
> > > > it. I would guess their (URI) brokeness has to do with
> > the DNS case
> > > > insensitivity (they didn't think anyone else would come up with
a
> > > > different way of naming authorities i bet you).
> > > >
> > > >     -Gabe
> > > > -----Original Message-----
> > > > From: Dave McAlpin [mailto:dave.mcalpin@epokinc.com]
> > > > Sent: Tuesday, October 21, 2003 12:36 PM
> > > > To: 'Wachob, Gabe'; 'Drummond Reed (E-mail)'; 'Lindelsee, Mike'
> > > > Subject: RE: case sensitivity of XRI authorities
> > > > In the previous discussion, we decided to leave it broken
> > > with respect
> > > > to internationalization because Nat said it was impossible
> > > to come up
> > > > with a generic, case insensitive comparison algorithm. At the
> > > > same time,
> > > > there was support for case insensitive comparison, so we
> > > > decided not to
> > > > throw the baby out with the bathwater and leave it enabled
> > > > for the ALPHA
> > > > production. Another way of looking at it is that we just
> > > > followed 2396's
> > > > lead on the authority portion.
> > > >
> > > > -----Original Message-----
> > > > From: Wachob, Gabe [mailto:gwachob@visa.com]
> > > > Sent: Tuesday, October 21, 2003 12:31 PM
> > > > To: Dave McAlpin (E-mail); Drummond Reed (E-mail); Lindelsee,
Mike
> > > > Subject: case sensitivity of XRI authorities
> > > >
> > > > What was the resolution on this topic?
> > > >
> > > > We have the language in there about case insensitive for alpha
> > > > characters.
> > > >
> > > > Is this what we concluded? I can't find a record in email.
> > > >
> > > > This is a little odd, of course, because for languages
> > like Spanish,
> > > > xri:+pequeño and xri:+Pequeño are the same but xri:+pequeño and
> > > > xri:+PEQUEÑO are different. That smacks of indifference to
> > > > internationalization concerns (its odd because whether or not
the
> > > > upcasing/downcasing changes equivalence depends on which
> > > > characters you
> > > > use).
> > > >
> > > > My vote is against case insensitivity even in the
> > > "authority" part, at
> > > > least as a rule that applies to all xris. We could say that, for
> > > > example, within the + namespace, the naming authorities are case
> > > > insensitive (with whatever definition of case insensitivity
> > > we decide
> > > > makes sense), but it strikes me as broken to declare case
> > > > insensitivity
> > > > the way we have done it.
> > > >
> > > >     -Gabe
> > > >
> > > >
> > > >
> > >
> >
>
>
>