Re: Dtd, Xsd or RNG/RNC as your choice of schema technology foryour docset?

[Norman Walsh <ndw@nwalsh.com>]

/ "Rajal Shah" <rajal@meshsoftware.com> was heard to say:
| I have a question for the community with regards to XML based authoring.. In
| a hypothetical scenario, if you were to redo all your documentation in XML
| in your enterprise (lets consider we create a custom book format - that
| maybe extended from docbook).. which schema technology would you choose to
| define your template rules - DTD, XSD or RNG/RNC format?

Unless there were compelling arguments in favor of using XSD, I'd use
RNG. Even if there were compelling arguments, I might very well use
RNG and create the XSD with trang. DTDs are dead.

| The main issues I want to consider in making the decision is:
|
| 1.	Authoring tools available. Cause non-technical folks would be
| authoring content and thus authoring tools need to support the schema
| technology we use.

I think it's a pretty mixed bag at this stage. There are some tools
that support XSD, there are some tools that support RNG, there are
some tools that support DTDs. Maybe the "big names" aren't supporting
RNG (yet), but the folks who are supporting it are doing a really nice
job. (And some of the XSD tools I encounter with unfortunate
regularity are demonstrably broken.)

| 2.	Content Management Repository and Storage requirements. (I don't see
| schema technology being too much of an issue here, except possibly
| validation content that is being checked-in/checked-out).

Yeah, if you're building an enterprise scale system, you'll be able to
configure it. Even for smaller systems, you can probably work around the
issue.

| 3.	Developing publishing tools.. (mostly XSL, so I don't think schema
| technology is an issue as long as the instance document validates with the
| schema).

Yep. XSLT2 will give you some data types, but I hope to see RELAX NG
validators that build XPath2 Data Models.

| My current thoughts are:
|
| 1.	I'd rather use RNG/RNC technology as my docset will exist for
| several years and would thus benefit in adopting newer technologies at the
| start..

It's not so much newer as better :-). Seriously, XSD may be great for
some applications (I really try not to bash it too hard) but few of
it's design strengths play towards constraining documents that contain
a lot of mixed content and recursive structure.

| 2.	However RNG/RNC has very limited support in GUI based authoring
| tools to help tech-writers, I think..

Growing every day :-)

| 3.	I could use XSDs: One big issue I had with XSDs were that we could
| not define our own entities  e.g. "&compName;" cannot be defined to output
| the Company Name consistently.. And we really need this functionality..
| (xinclude may be the answer)..

You can't do that with RNG either.

But you can use internal subsets without using DTDs. And if you don't
need the macros to appear in attribute values, you can even use
elements. (<m:compName/> is expanded by your processing system.)

| 4.	Also with XSDs we can't use the internal DOCTYPE for each document
| and thus overwrite some of the external entity declaration in each
| instance..

That's usually considered a feature. Allowing documents to dynamically
change the schema that they validate against can be problematic. But
you can't do that with XSD or RNG. If you need that, you're stuck with
DTDs.

| 5.	DTD is safest in some sense, but all new enhancements are being done
| in XSD and RNG/RNC technologies only.. So it's not the best idea to start a
| new docset with DTDs anymore.

I agree.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com>      | There is a great difference
http://www.oasis-open.org/docbook/ | between seeking how to raise a
Chair, DocBook Technical Committee | laugh from everything, and seeking
                                   | in everything what may justly be
                                   | laughed at.--Lord Shaftesbury

PGP signature