ambiguous content models

[LynnHales]

Epic is one of the few XML applications that still reports 'ambiguous'
content models as an error when parsing a DTD (yes some of us still use
them). I have a DTD that the requirements document is forcing a change to.
The change is causing Epic to report the model as ambiguous. It took me a
while to see where the ambiguity lies, but finding a 'fix' has been elusive.
My boss and I have both looked at this for a couple of hours (took him a bit
longer to see the ambiguous content).

I'm going to continue looking, but here is the model. I'm open to
suggestions if anyone feels like jumping in. 😄

(interconnect | testflow | funcdepend | schematic | comp-locator |
harness-indx)*), proc)*, (sysdesc?, (interconnect | testflow | funcdepend |
schematic | comp-locator | harness-indx)*), (hookup?, tsproc, disconnect?),
followon.maintsk?)>

The group (sysdesc?, (interconnect | testflow | funcdepend | schematic |
comp-locator | harness-indx)*) is declared as a parameter ENTITY in the DTD.


Epic reports the element <interconnect> is used ambiguously. What astounds
me is why the <sysdesc> is not reported as ambiguous as it falls into the
same criteria as the <interconnect>.

Many thanks

Lynn

[GregoryMackenzi]

Hi Lynn,

I'm wondering about the construction of the element itself since sometimes the first thing reported is not always the issue. For example, I wonder if interconnect were repositioned in the sequence or removed from the model if it would report testflow as ambiguous since it would then be first, something like:

(sysdesc?, ( testflow | interconnect | funcdepend | schematic | comp-locator | harness-indx)*)

Greg

[ClayHelberg]

Hi Lynn--

One thing that jumps out at me, is this: if I'm reading this right, it looks like your main content section (the entity refs with "(sysdesc?,...)") can legally be empty. In other words, a document could go straight from <intro> to <hookup> (or even end right after <initial-setup>. Is that what you intend?

--Clay

[LynnHales]

Well my boss found the solution.



He grouped the <proc> with the second rather than the first group with the
<interconnect>. The proc is then required between the two ambiguous
elements.



Lynn

[LynnHales]

Clay,



Yes, we tried to clean it up, but the users want to everything, so it was
back to the old way only allow it to repeat. L



Lynn

[ClayHelberg]

OK, just checking. (I feel your pain--negotiating structure vs. freedom with authors can be dicey.)