Re: CPIM and URIs

At 02:08 PM 5/24/02 -0400, John D. Ramsdell wrote:
>The one pain in the neck about this is one cannot use Java 2 Standard
>Edition 1.4's java.net.URI class to do the work for you since you have
>to keep track of which '@' separates the local part and the domain.
>Maybe the scheme specific part should simply be an encoding of the
>external representation of an Email address.

[Full message below, for context]

??

You've lost me now.  The syntax I mentioned allows exactly one '@' in an 
im: URI.

Further, for the purposes of URI handling, the scheme-specific part 
*should* be opaque, IMO.  (I don't think it's reasonable to expect generic 
URI code to disassemble an application-dependent format.)

What am I missing here?

#g
--

>Graham Klyne <[email protected]> writes:
>
> > But I'm worried that this is making a simple thing complicated, or at
> > least difficult to follow for people who aren't familiar with the
> > other related specs.  Maybe, instead, we could just define a syntax
> > for an im: URI, and leave it at that?
>
>I think the definition I gave should be abandoned for two reasons.
>One is that I like your suggestion better, but the other reason is
>that the definition I gave was ambiguous.  There was no way to
>identify the '@' character that separates the local part from the
>domain.  After removing the invisible characters from the following
>two examples, tell me how to separate the local part from the domain.
>
>      a@[@\[@]
>
>      "a@]"@[@]
>
> > E.g. something like this
> >
> > [[
> >     Inboxes are named using URIs with the following ABNF syntax:
> >
> >         ;; Domain is defined in [RFC2821], either a FQDN or a literal
> >         inbox       = "im:" local "@" Domain
> >
> >         local       = address [ "/" subaddress ]
> >
> >         address     = token
> >
> >         subaddress  = token
> >
> >         ;; all non-control characters, excluding "/" and "@" delimiters
> >         token       = 1*(%x20-2E / %x30-3F / %x41-7E / UTF-8) ;; [RFC2044]
> >
> >     URI character escaping (%hh) is used as necessary when presenting the
> >     URI to an environment that expects a US-ASCII URI character sequence
> >     to be used [RFC2396, section 2.1].
> > ]]
> >
> > (This was adapted from email addresses, but is not identical to
> > RFC2821/2 syntax.)
>
>The one pain in the neck about this is one cannot use Java 2 Standard
>Edition 1.4's java.net.URI class to do the work for you since you have
>to keep track of which '@' separates the local part and the domain.
>Maybe the scheme specific part should simply be an encoding of the
>external representation of an Email address.
>
>   URI u is an instant inbox URI if and only if all of the following
>   are true:
>
>         * The scheme of u is the string "im".
>
>         * The decoded scheme specific part of u is an addr-spec as
>           defined by the non-obsolete productions given in RFC 2822.
>
>         * The fragment part of u is absent.
>
>On the other hand, writing a parser for your productions seems
>straightforward.
>
>John

-------------------
Graham Klyne
<[email protected]>



  [reminder: [email protected] for non-technical discussions, please]