odd -> xsd fault?

classic Classic list List threaded Threaded
11 messages Options
Reply | Threaded
Open this post in threaded view
|

odd -> xsd fault?

Eduard Drenth
Dear all,

When generating xsd from odd using 4.0 stylesheets from latest oxygen, I get this incorrect definition:

   <xs:element name="taxonomy">
      <xs:annotation>
         <xs:documentation>defines a typology either implicitly, by means of a bibliographic citation, or explicitly by a structured taxonomy. [2.3.7. The Classification Declaration]</xs:documentation>
      </xs:annotation>
      <xs:complexType>
         <xs:choice>
            <xs:choice>
               <xs:choice maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
               <xs:choice minOccurs="0" maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
            </xs:choice>
            <xs:choice minOccurs="0" maxOccurs="unbounded">
               <xs:element ref="tei:category"/>
               <xs:element ref="tei:taxonomy"/>
            </xs:choice>
         </xs:choice>
      </xs:complexType>
   </xs:element>

The source of this problem is:

                <moduleRef key="header" source="https://www.tei-c.org/release/xml/tei/odd/p5subset.xml"/>

Could there be something wrong in the stylesheet or p5subset?

Regards, Eduard

-- 

Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder


lou
Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

lou
Most peculiar. This content model is missing elements not defined in the header module, which leads me to suspect that your ODD does not include for example &lt;desc> or &lt;gloss>. Can you confirm what other modules or elements are referenced in your ODD?


On Sat, 1 Aug 2020 at 13:43, Eduard Drenth <[hidden email]> wrote:
Dear all,

When generating xsd from odd using 4.0 stylesheets from latest oxygen, I get this incorrect definition:

   <xs:element name="taxonomy">
      <xs:annotation>
         <xs:documentation>defines a typology either implicitly, by means of a bibliographic citation, or explicitly by a structured taxonomy. [2.3.7. The Classification Declaration]</xs:documentation>
      </xs:annotation>
      <xs:complexType>
         <xs:choice>
            <xs:choice>
               <xs:choice maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
               <xs:choice minOccurs="0" maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
            </xs:choice>
            <xs:choice minOccurs="0" maxOccurs="unbounded">
               <xs:element ref="tei:category"/>
               <xs:element ref="tei:taxonomy"/>
            </xs:choice>
         </xs:choice>
      </xs:complexType>
   </xs:element>

The source of this problem is:

                <moduleRef key="header" source="https://www.tei-c.org/release/xml/tei/odd/p5subset.xml"/>

Could there be something wrong in the stylesheet or p5subset?

Regards, Eduard

-- 

Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth

Op freed bin ik thús/wurkje ik minder


Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Bauman, Syd
In reply to this post by Eduard Drenth
Sorry … I sent this to OP directly, when I intended the list.

Thus your reply came directly to me, Eduard, and Lou did not get the benefit of seeing your ODD. However, I do not get the benefit of seeing it either, because Outlook is apparently convinced that it is dangerous and I should not be allowed to.

Can you re-send it as a ZIP or tar archive, or as a .txt file or some such? Either to me or the list, your preference.


Others may have more intelligent answers, but I would need more information to say much useful. What is the actual error or problem with that snippet of XSD? (I do not read XSD very well.) I am guessing that the error is non-determinism, but there might be something wrong that flags an error before the processor notices that the content model is non-deterministic.

Is there a reason you are using XSD not RNG (or RNC, same thing)? IIRC, the build process generates a RELAX NG schema, which gets converted to XSD using `trang`. We test the RELAX NG much more thoroughly than the XSD or DTD schemas.

So the short answer for you may be to switch to RELAX NG (which does not require determinism). If that solves your problem, then the longer answer for the TEI is that we have a bug in schema generation.

Can you post the complete ODD that generates the problem? What processor are you using?



When generating xsd from odd using 4.0 stylesheets from latest oxygen, I get this incorrect definition:

   <xs:element name="taxonomy">
      <xs:annotation>
         <xs:documentation>defines a typology either implicitly, by means of a bibliographic citation, or explicitly by a structured taxonomy. [2.3.7. The Classification Declaration]</xs:documentation>
      </xs:annotation>
      <xs:complexType>
         <xs:choice>
            <xs:choice>
               <xs:choice maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
               <xs:choice minOccurs="0" maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
            </xs:choice>
            <xs:choice minOccurs="0" maxOccurs="unbounded">
               <xs:element ref="tei:category"/>
               <xs:element ref="tei:taxonomy"/>
            </xs:choice>
         </xs:choice>
      </xs:complexType>
   </xs:element>

The source of this problem is:

                <moduleRef key="header" source="https://www.tei-c.org/release/xml/tei/odd/p5subset.xml"/>

Could there be something wrong in the stylesheet or p5subset?

Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Eduard Drenth
In reply to this post by Eduard Drenth
My original reply, now to the list as well....with odd as txt attached

Hi,

I use xsd because I know it better and do a lot of java, there is more standardized support for xsd such as jaxb. Also tools like intellij, oxygen, netbeans, exide directly work with it,
not only for validation, but also for documentation.

I may one day switch to rng thow.

So here is what I do:

1) compile linguistic_attributes.odd into linguistic_attributes.odd.compiled using latest oxygen odd2odd.xsl
2) transform tdb.odd into tdb.xsd using latest oxygen tei odd to schema

see attachment

Bye, Eduard


-----Original Message-----
From: "Bauman, Syd" <[hidden email]>
To: Eduard Drenth <[hidden email]>
Subject: Re: odd -> xsd fault?
Date: Sat, 01 Aug 2020 14:09:45 +0000

Others may have more intelligent answers, but I would need more information to say much useful. What is the actual error or problem with that snippet of XSD? (I do not read XSD very well.) I am guessing that the error is non-determinism, but there might be something wrong that flags an error before the processor notices that the content model is non-deterministic.

Is there a reason you are using XSD not RNG (or RNC, same thing)? IIRC, the build process generates a RELAX NG schema, which gets converted to XSD using `trang`. We test the RELAX NG much more thoroughly than the XSD or DTD schemas.

So the short answer for you may be to switch to RELAX NG (which does not require determinism). If that solves your problem, then the longer answer for the TEI is that we have a bug in schema generation.

Can you post the complete ODD that generates the problem? What processor are you using?



When generating xsd from odd using 4.0 stylesheets from latest oxygen, I get this incorrect definition:

   <xs:element name="taxonomy">
      <xs:annotation>
         <xs:documentation>defines a typology either implicitly, by means of a bibliographic citation, or explicitly by a structured taxonomy. [2.3.7. The Classification Declaration]</xs:documentation>
      </xs:annotation>
      <xs:complexType>
         <xs:choice>
            <xs:choice>
               <xs:choice maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
               <xs:choice minOccurs="0" maxOccurs="unbounded">
                  <xs:element ref="tei:category"/>
                  <xs:element ref="tei:taxonomy"/>
               </xs:choice>
            </xs:choice>
            <xs:choice minOccurs="0" maxOccurs="unbounded">
               <xs:element ref="tei:category"/>
               <xs:element ref="tei:taxonomy"/>
            </xs:choice>
         </xs:choice>
      </xs:complexType>
   </xs:element>

The source of this problem is:

                <moduleRef key="header" source="https://www.tei-c.org/release/xml/tei/odd/p5subset.xml"/>

Could there be something wrong in the stylesheet or p5subset?

-- 

Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder



-- 

Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder



-- 
Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder




tdb.odd.txt (2K) Download Attachment
linguistic_attributes.odd.txt (46K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Bauman, Syd
In reply to this post by Eduard Drenth
We can discuss the advantages and disadvantages of RNG vs XSD some other time, perhaps. In the meantime, your main ODD (linguistic_attributes.odd) is *almost* there.

Rather than “compiling” it with odd2odd.xsl (which is intended to be one part of a larger transformation, not really to be used on its own), using the “TEI ODD to XML Schema” transformation scenario is probably your best bet. It will just read in linguistic_attributes.odd and write out a usable XSD, although the ODD file needs a minor fix, first. (I am not sure if the output is called out/linguistic_attributes.xsd, out/linguistic_attributes.odd.xsd, or out/linguistics.xsd, as I almost never use oXygen for this purpose.[1])

To get your linguistics_attributes.odd file working all I did was add two lines:
         <moduleRef key="header"/>
         <moduleRef key="textstructure"/>
I tested that it produced a valid XSD. I tested that the schema works using the output Relax NG (because I am clumsy with XSD.)

Notes
[1] I typically used the commandline version of the same stylesheets.


My original reply, now to the list as well....

Hi,

I use xsd because I know it better and do a lot of java, there is more standardized support for xsd such as jaxb. Also tools like intellij, oxygen, netbeans, exide directly work with it,
not only for validation, but also for documentation.

I may one day switch to rng thow.

So here is what I do:

1) compile linguistic_attributes.odd into linguistic_attributes.odd.compiled using latest oxygen odd2odd.xsl
2) transform tdb.odd into tdb.xsd using latest oxygen tei odd to schema

see attachment

Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Eduard Drenth
Well, I don't think so, tdb.odd is chained with linguistic_attributes.odd.

So I compile linguistic_attributes.odd and than generate xsd from tdb.odd.

In the mean time I got rid of the faulty tdb.xsd by setting up my odds a bit different, see attachments.

Tricky stuff, this chaining, my whole life depends on it, so I hope it will be maintained...

Bye, Eduard


-----Original Message-----
From: "Bauman, Syd" <[hidden email]>
Reply-To: "Bauman, Syd" <[hidden email]>
Subject: Re: odd -> xsd fault?
Date: Sat, 01 Aug 2020 19:50:29 +0000

We can discuss the advantages and disadvantages of RNG vs XSD some other time, perhaps. In the meantime, your main ODD (linguistic_attributes.odd) is *almost* there.

Rather than “compiling” it with odd2odd.xsl (which is intended to be one part of a larger transformation, not really to be used on its own), using the “TEI ODD to XML Schema” transformation scenario is probably your best bet. It will just read in linguistic_attributes.odd and write out a usable XSD, although the ODD file needs a minor fix, first. (I am not sure if the output is called out/linguistic_attributes.xsd, out/linguistic_attributes.odd.xsd, or out/linguistics.xsd, as I almost never use oXygen for this purpose.[1])

To get your linguistics_attributes.odd file working all I did was add two lines:
         <moduleRef key="header"/>
         <moduleRef key="textstructure"/>
I tested that it produced a valid XSD. I tested that the schema works using the output Relax NG (because I am clumsy with XSD.)

Notes
[1] I typically used the commandline version of the same stylesheets.


My original reply, now to the list as well....

Hi,

I use xsd because I know it better and do a lot of java, there is more standardized support for xsd such as jaxb. Also tools like intellij, oxygen, netbeans, exide directly work with it,
not only for validation, but also for documentation.

I may one day switch to rng thow.

So here is what I do:

1) compile linguistic_attributes.odd into linguistic_attributes.odd.compiled using latest oxygen odd2odd.xsl
2) transform tdb.odd into tdb.xsd using latest oxygen tei odd to schema

see attachment

-- 
Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder




linguistic_attributes.odd.txt (46K) Download Attachment
tdb.odd.txt (2K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Bauman, Syd
Well, yes, but tdb.odd does not do anything (except add some modules) differently than linguistic_attributes.odd, so I (apparently incorrectly) presumed the desired output was a TEI schema with the att.features added.

Given that tdb.odd does not do anything, I do not see why you want to chain these two ODDs, but my suspicion is that you … oh, looks like you have already discovered that you need (at least) the 4 required modules in linguistic_attributes. It is not clear to me why one would have to replicate the <moduleRef>s from linguistic_attributes in tdb, but it does seem you do. In any case, it is probably easier and more robust, although some would argue a little less obvious and precise, to use "tei:current" instead of the full URL to the current p5subset in the Vault.

So, what is it you are trying to accomplish (i.e., why is chaining so important)? Not that anyone is planning to remove chaining capability, but it is somewhat tenuous, IMHO.


Well, I don't think so, tdb.odd is chained with linguistic_attributes.odd.

So I compile linguistic_attributes.odd and than generate xsd from tdb.odd.

In the mean time I got rid of the faulty tdb.xsd by setting up my odds a bit different, see attachments.

Tricky stuff, this chaining, my whole life depends on it, so I hope it will be maintained...


Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Eduard Drenth
Perhaps you are right, I'll review all the code generation I have that uses linguistic_attributes.odd and see if I can do without chaining.

One reason for having linguistic_attributes separately is that I like to modularize and intended linguistic_attributes.odd to be usefull to others.
But since I need to apply several tricks (like repeating moduleRef) to get chaining to produce valid (though I am not even sure it does)
and usable output and since nobody uses linguistic_attributes.odd as far as I know....

Thanks for thinking along, Eduard

-----Original Message-----
From: "Bauman, Syd" <[hidden email]>
Reply-To: "Bauman, Syd" <[hidden email]>
Subject: Re: odd -> xsd fault?
Date: Sat, 01 Aug 2020 22:39:34 +0000

Well, yes, but tdb.odd does not do anything (except add some modules) differently than linguistic_attributes.odd, so I (apparently incorrectly) presumed the desired output was a TEI schema with the att.features added.

Given that tdb.odd does not do anything, I do not see why you want to chain these two ODDs, but my suspicion is that you … oh, looks like you have already discovered that you need (at least) the 4 required modules in linguistic_attributes. It is not clear to me why one would have to replicate the <moduleRef>s from linguistic_attributes in tdb, but it does seem you do. In any case, it is probably easier and more robust, although some would argue a little less obvious and precise, to use "tei:current" instead of the full URL to the current p5subset in the Vault.

So, what is it you are trying to accomplish (i.e., why is chaining so important)? Not that anyone is planning to remove chaining capability, but it is somewhat tenuous, IMHO.


Well, I don't think so, tdb.odd is chained with linguistic_attributes.odd.

So I compile linguistic_attributes.odd and than generate xsd from tdb.odd.

In the mean time I got rid of the faulty tdb.xsd by setting up my odds a bit different, see attachments.

Tricky stuff, this chaining, my whole life depends on it, so I hope it will be maintained...


-- 
Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder



lou
Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

lou
Chaining of ODDs is ideal in the situation where you have a functioning (complete) customization which you want to modify further, typically to subset it in different ways, or (as in your TBD case) to add extra TEI bits. Your use case, if I understand it correctly, is slightly different: you have a set of declarations (your linguistic attributes) which you want to plug into a variety of schemas. Maybe one day the ODD system will be enhanced to allow you to define and include your own module alongside the TEI defined ones, but that can't readily be done now. You could probably do it by defining your linguistic attributes in RNG and then using <moduleRef> to pull in the RNG code directly, but that would be a vile hack, even though it's what the TEI recommends for embedding mathml.

Have you considered using xInclude to embed the source code for your attribute declarations in each ODD? That should work, and would enable you to maintain the linguistic attributes in a single separate document.

The other interesting question (well, it interests me) raised by your query is how best to alert ODD authors to the kind of inter-element dependency implied by some content models. With no member of either model.descLike or model.descLike included in a schema, the content model for &lt;taxonomy> is irreparably indeterminate, which is almost certainly not what was intended. And there are almost certainly  other similar cases, but it would be a challenge to track them down except by inspection...


 

On Sun, 2 Aug 2020 at 09:32, Eduard Drenth <[hidden email]> wrote:
Perhaps you are right, I'll review all the code generation I have that uses linguistic_attributes.odd and see if I can do without chaining.

One reason for having linguistic_attributes separately is that I like to modularize and intended linguistic_attributes.odd to be usefull to others.
But since I need to apply several tricks (like repeating moduleRef) to get chaining to produce valid (though I am not even sure it does)
and usable output and since nobody uses linguistic_attributes.odd as far as I know....

Thanks for thinking along, Eduard

-----Original Message-----
From: "Bauman, Syd" <[hidden email]>
Reply-To: "Bauman, Syd" <[hidden email]>
Subject: Re: odd -> xsd fault?
Date: Sat, 01 Aug 2020 22:39:34 +0000

Well, yes, but tdb.odd does not do anything (except add some modules) differently than linguistic_attributes.odd, so I (apparently incorrectly) presumed the desired output was a TEI schema with the att.features added.

Given that tdb.odd does not do anything, I do not see why you want to chain these two ODDs, but my suspicion is that you … oh, looks like you have already discovered that you need (at least) the 4 required modules in linguistic_attributes. It is not clear to me why one would have to replicate the <moduleRef>s from linguistic_attributes in tdb, but it does seem you do. In any case, it is probably easier and more robust, although some would argue a little less obvious and precise, to use "tei:current" instead of the full URL to the current p5subset in the Vault.

So, what is it you are trying to accomplish (i.e., why is chaining so important)? Not that anyone is planning to remove chaining capability, but it is somewhat tenuous, IMHO.


Well, I don't think so, tdb.odd is chained with linguistic_attributes.odd.

So I compile linguistic_attributes.odd and than generate xsd from tdb.odd.

In the mean time I got rid of the faulty tdb.xsd by setting up my odds a bit different, see attachments.

Tricky stuff, this chaining, my whole life depends on it, so I hope it will be maintained...


-- 
Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth

Op freed bin ik thús/wurkje ik minder



Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Bauman, Syd
Have you considered using xInclude to embed the source code for your attribute declarations in each ODD? That should work, and would enable you to maintain the linguistic attributes in a single separate document.

Or <specGrpRef>?

Reply | Threaded
Open this post in threaded view
|

Re: odd -> xsd fault?

Eduard Drenth
In reply to this post by lou
Again posted to the list, now without rejected pgp signature...

Your description of my use-case is quite correct. I don't know if the ability to plugin a custon module will be of any help for me, all I need is to
restrict @tei:pos and to add @fa:person etc. and than to have those available on tei:w, tei:join and tei:m.

My own description of my use-cases:

1) Corpus linguistics: I want to use a tightened set of linguistic attributes for both the editing process (oxygen) and doing research (blacklab)
2) Dictionaries: I want to restrict the contents of tei:gram to the same set of linguistics (pos.noun, person.first, etc.), different, far more restricted odd here
3) Lexicons (word lists): I want to use the same set of linguistics again
4) Digital editions: I want to use the same dataformat as for corpus linguistics and use TEI processing model and teipublisher for publication

This enhances integration possibilities, enables me to maintain sources in one place and to benefit from and contribute to the TEI, UD and other communities.

xinclude may be of help yes.

Your last remark I do not quite follow.

- Does "inter-element dependency" between models mean that one model needs another, so depenedncy management (like maven)?
- Would like to enhance support for odd authors like me in there editing process?
- "not what was intended" by me?
- Do you miss some info in my odd's?
- What you put in model.descLike to track inter-element dependency? Plain text? Something that enables automatic dependency checking?

Well, perhaps I just miss some background here... Anyway I do think, judging from my experience, that odd development could do with better guidance.
And I also do think odd deserves a more prominent place in the TEI, among other things because it is a reliable basis for tool development.

Bye, Eduard

-----Original Message-----
From: Lou Burnard <[hidden email]>
To: Eduard Drenth <[hidden email]>
Cc: Liste TEI <[hidden email]>
Subject: Re: odd -> xsd fault?
Date: Sun, 02 Aug 2020 09:48:24 +0100

Chaining of ODDs is ideal in the situation where you have a functioning (complete) customization which you want to modify further, typically to subset it in different ways, or (as in your TBD case) to add extra TEI bits. Your use case, if I understand it correctly, is slightly different: you have a set of declarations (your linguistic attributes) which you want to plug into a variety of schemas. Maybe one day the ODD system will be enhanced to allow you to define and include your own module alongside the TEI defined ones, but that can't readily be done now. You could probably do it by defining your linguistic attributes in RNG and then using <moduleRef> to pull in the RNG code directly, but that would be a vile hack, even though it's what the TEI recommends for embedding mathml.

Have you considered using xInclude to embed the source code for your attribute declarations in each ODD? That should work, and would enable you to maintain the linguistic attributes in a single separate document.

The other interesting question (well, it interests me) raised by your query is how best to alert ODD authors to the kind of inter-element dependency implied by some content models. With no member of either model.descLike or model.descLike included in a schema, the content model for &lt;taxonomy> is irreparably indeterminate, which is almost certainly not what was intended. And there are almost certainly  other similar cases, but it would be a challenge to track them down except by inspection...


 

On Sun, 2 Aug 2020 at 09:32, Eduard Drenth <[hidden email]> wrote:
Perhaps you are right, I'll review all the code generation I have that uses linguistic_attributes.odd and see if I can do without chaining.

One reason for having linguistic_attributes separately is that I like to modularize and intended linguistic_attributes.odd to be usefull to others.
But since I need to apply several tricks (like repeating moduleRef) to get chaining to produce valid (though I am not even sure it does)
and usable output and since nobody uses linguistic_attributes.odd as far as I know....

Thanks for thinking along, Eduard

-----Original Message-----
From: "Bauman, Syd" <[hidden email]>
Reply-To: "Bauman, Syd" <[hidden email]>
Subject: Re: odd -> xsd fault?
Date: Sat, 01 Aug 2020 22:39:34 +0000

Well, yes, but tdb.odd does not do anything (except add some modules) differently than linguistic_attributes.odd, so I (apparently incorrectly) presumed the desired output was a TEI schema with the att.features added.

Given that tdb.odd does not do anything, I do not see why you want to chain these two ODDs, but my suspicion is that you … oh, looks like you have already discovered that you need (at least) the 4 required modules in linguistic_attributes. It is not clear to me why one would have to replicate the <moduleRef>s from linguistic_attributes in tdb, but it does seem you do. In any case, it is probably easier and more robust, although some would argue a little less obvious and precise, to use "tei:current" instead of the full URL to the current p5subset in the Vault.

So, what is it you are trying to accomplish (i.e., why is chaining so important)? Not that anyone is planning to remove chaining capability, but it is somewhat tenuous, IMHO.


Well, I don't think so, tdb.odd is chained with linguistic_attributes.odd.

So I compile linguistic_attributes.odd and than generate xsd from tdb.odd.

In the mean time I got rid of the faulty tdb.xsd by setting up my odds a bit different, see attachments.

Tricky stuff, this chaining, my whole life depends on it, so I hope it will be maintained...


-- 

Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth

Op freed bin ik thús/wurkje ik minder



-- 

Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder



-- 
Eduard Drenth, Software Architekt


Doelestrjitte 8
8911 DX  Ljouwert
+31 58 234 30 47
+31 62 094 34 28 (privé)

skype: eduarddrenth
frisian.eu

Op freed bin ik thús/wurkje ik minder