geekodoc only: Restrict IDs to only alphanumeric characters + "-" about geekodoc

Fwiw, there is now a Travis check against such IDs, so most of our documents are now checked for this.

Maybe it is not such a good idea to do that in the schema and the Travis-based solution is better suited for this issue anyway because it is less fix than the schema and because we'd immediately invalidate a ton of existing documents that validated before.

I almost think we should close this one. @tomschr ?

[tomschr]

@sknorr

I can understand that it gives mixed feelings (because it invalidates valid documents). Same with me. 😉 Mabye we should discuss this personally. I provide some thoughts.

Nevertheless, I think it is a good idea. I run into the trap with one of my files and asking myself what I did wrong. I found it later in Travis. IMHO, I think, we shouldn't have two validations which exists in parallel: a local one and a remote one in Travis. IMHO both must report the same (there may be valid exception to this rule). So faster reporting of such errors is a good thing.

On the other side, I see the problem with former documents. With this change they may be invalid now. This is probably a bad thing. However, we could do:

• Raise the major version of Geekodoc to indiciate, there is an incompatible change.
• Communicate the change on doku-intern, get feedback, raise awareness etc.
• Write a conversation script (XSLT, bash, ...) which converts all attribute values of xml:id and linkend. Didn't you do that already?

We could also deal with this change on a packaging level. For example, we could introduce a geekodoc2 package which could co-exist with the (now) old geekodoc. Same like with DocBook 4.x and 5.x. They all co-exist peacefully.

I would suggest, let's discuss this in person (faster, easier). 👍

[tomschr]

Revisiting this issue in the light of openSUSE/daps#566. To use the ID checks in Travis this can be replicated in GeekoDoc as well.
Tested it with this snippet of RNC code:

include "docbookxi.rnc"
{
   db.xml.id.attribute =
     ## Identifies the unique ID value of the element
     attribute xml:id {
      xsd:ID  { pattern = "[\-a-zA-Z0-9]+" } }
}

It validates an xml:id of foo-bar successfully with jing, but not foo.bar.

That is imo definitely a better solution than using DAPS itself for this purpose, thanks. Regex looks good too (though I wonder if the leading \ in the character set is necessary -- but maybe that is necessary because the RNG regex syntax is a bit weird).