ERB call on addressing
After *endless* further discussion, and realization that co-existing with
the web is hard, the ERB, on March 22, voted as follows:
A locator is a string which may contain either or both a URL
and a TEI extended pointer [Xptr]. The URL indicates a resource; if
the Xptr appears, this means that the desired resource is a
"sub-resource" of that indicated by the URL. The URL must appear first
in the locator string. If the URL does not appear, the Xptr is to
be applied to the current document. If an Xptr appears, it must be
preceded by a Separator character. There are three possible separator
# - means that the user-agent is to fetch the resource described by
the URL, and use the Xptr to extract the desired sub-resource.
? - means that the user-agent is to transmit the URL and Xptr to the
server, which is to use the Xptr to extract the desired sub-resource
and transmit it to the user-agent. In this case, the Xptr must
be preceded by the string "XML-PTR="
| - means that this locator only expresses the fact that the desired
sub-resource is to be retrieved by applying the Xptr to the resource
identified by the URL. No constraint is placed on the system as
to how this should be accomplished.
1. '#' and '?' are standard Web practice; the appropriate RFC's make it
clear that it would be very unwise for us to try to overload the
2. The "XML-PTR=" on the '?' form is an effort to allow this to work
smoothly with CGI gateways; while CGI is of diminishing importance,
it's an awfully easy way to set up a test-bed, and other more
efficient query protocols often hide behind CGI syntax anyhow
3. The choice of the '|' character for the role defined here is open
to debate. We wanted something that, per RFC, is supposed to be
escaped in URLs, so that its naked appearance is a signal that
something special is happening; but that isn't going to wreak too
much havoc. Possible alternatives are '^', '@', and '<'. Those
who'd like to debate this point MUST READ THE RFC FIRST!
4. The 'special' case where the item in the Xptr position is just a
a string (e.g. http://www.xml.com/faq.xml#A27 or some such) is no
longer a special case; we will define an interpretation of a
single unmarked string as a TEI Xpointer; we are already going to
have to mung the syntax of these anyhow in order to crush spaces...
the sense of the ERB is that anything that Michael and Steve
come up with, presumably in conclave with the TEI gang, will
probably be OK by us.
Cheers, Tim Bray
email@example.com http://www.textuality.com/ +1-604-708-9592