We are attempting to implement XML for
one of our projects.
There is a need for our users to cross reference
documents/data.
We were looking for a way to generate/format a universal
ID.
FPI seems to solve this issue but:
1- Why having yet another syntax ? I mean
-//myOrg//myDocType
myDocName//en
could be:
<Registered>No</Registered>
<Org>myOrg</Org>
<DocType>myDocType</DocType>
<DocName>myDocName</DocName>
<Lang>en</
Lang>
It
could be, but then you couldnt use it inside attributes (unless you put in a
processor to parse tags) or thePUBLIC identifier field of an entity
declaration (unless ditto).
The question is the same as "why aren't URLs marked
up using element syntax?" Various previous DTDs have tried to do this,
and they have not been as successful. I think the reason can be
explained by the software engineering concept of "cohesion and
coupling": some kinds of data so naturally cohere to each other
(i.e. in the minds of users) that it is wise to couple them (i.e. in their
syntax) strongly. In particular, some strings have certain qualities
where our minds accept them as names: I think URLs, FPIs and MIME types are
such.
In
my book on document patterns I discuss this. I would also mention that these
kind of compound names correspond to the "natural document"
pattern, in that they have ( metadata, data, annotations ). When there is
one natural document embedded in a different type of document (i.e. code
inside text, FPI inside attribute) it seems that humans appreciate this
being flagged: especially if no formatting is available, a change of
notation will be used.
Why
does CSS not use instance notation (apart from not upsetting dumb HTML
browsers)? I think it is exactly this reason that an embedded document of a
completely different document type (as far as its function is
concerned) is best marked up by clearly flagging it by using a
differnet notation.
In other words
<a
href="http://www.w3.org/TR/">
is not the same as
<a href:method="http" href:node="www.w3.org"
href:directory="TR/" >
but
<a
href:method="http" href:node="www.w3.org"
href:directory="TR/" href:combine-notation="URL
Syntax">
I
think XMLers should not go too far in saying that instance syntax is always
preferable to little languages (embedded notations), in particular when the
function of the embedded language is not structural markup but for naming or
locating or stylesheets.
2- Where can I find more info about FPIs ? I spend a long time with
a
web search engine but got only few interesting documents. Where can I
find ISO 9070 related documents ? The ISO web site is not very user
friendly
for a first timer.
Several (most?) SGML books have information on FPI syntax.
Online try Robin Cover's SGML site first.
3- Assuming that we keep this syntax, would it be
valid to have something
like:
-//anOrg::anAuthor//aDocType
aDocName::aVersionNumber//en
or even worse:
-//aCountry::anOrg::aDepartment::anAuthor//aDocType::aDocTypeVersionNumber
aDocName::aDocVersionNumber//en
After the "-//" you can use any convention you want for the
"owner of FPI" and "name" field. the syntax
is
-//owner of FPI//type
name//lang
where the "type" is defined in the SGML standard: e.g. DTD,
ENTITIES, NOTATION, TEXT, NONSGML (capitals) and the "lang"
(capitals) is the language code of the FPI (and therefore probably of the
resource). Why "even worse"? If that is what is
required to specify it exactly enough, that is great.
ISBN
and IDN (Internet Domain Name) "owners" may use the
form
+//ISBN number//type
name//lang
+//IDN addess//type
name//lang
Rick
Jelliffe
