xml SYSTEM vs PUBLIC

[Razvan]

Hi !

In an XML DOCTYPE declaration you can use either PUBLIC or
SYSTEM attributes:

<!DOCTYPE RootElement SYSTEM "SomeDTD.dtd">
<!DOCTYPE RootElement PUBLIC "SomeDTD.dtd">

When should I use SYSTEM and when should I use PUBLIC ?

Regards,
Razvan

 

[David Carlisle]

Hi !

In an XML DOCTYPE declaration you can use either PUBLIC or
SYSTEM attributes:

<!DOCTYPE RootElement SYSTEM "SomeDTD.dtd">
<!DOCTYPE RootElement PUBLIC "SomeDTD.dtd">

When should I use SYSTEM and when should I use PUBLIC ?

Regards,
Razvan

No, you could do that in SGML but in XML you _Always_ have to have a SYSTEM identifier. If you use PUBLIC you gave to give _both_ a public ID and the SYSTEM.

Although XML doesn't enforce this PUBLIC identifiers usually follow the Formal Public Identifier syntax which has a bunch of fields separated by / that encode the author and identity of the dtd file but not (usually0 its location. eg MathML uses:

<!DOCTYPE math
PUBLIC "-//W3C//DTD MathML 2.0//EN"
"http://www.w3.org/Math/DTD/mathml2/mathml2.dtd" The PUBLIC Identifier "-//W3C//DTD MathML 2.0//EN shows it's (not from ISO, the -) is from W3C is called DTD MathML 2.0 and uses English as its documentaion, (one possible) URI where you might actually find the thing is given by the SYSTEM ID "http://www.w3.org/Math/DTD/mathml2/mathml2.dtd"

David

 

[Razvan]

Hi David,

Thanks for your answer.

No, you could do that in SGML but in XML you _Always_ have to have a SYSTEM identifier.

If that is true then why have a PUBLIC identifier at all ? What is their purpose ?

Is there some limitation to the type of data that you can put in SYSTEM vs. PUBLIC identifiers ? For example: SYSTEM only supports URLs (like http://www.w3c.org/ ) while PUBLIC only supports local paths ?
(this is just an example)

Regards,
Razvan

 

[David Carlisle]

If that is true then why have a PUBLIC identifier at all ? What is their purpose

beware that you are skirting on the edge of a very deep "debate" Basically the public identifier gives a unique canonical name for the dtd that you can recognise; whereas the SYSTEM identifier may well point at a local path on the original author's system that doesn't make sense on yours. An XML parser may use either (or both) identifiers to find the file.

So if you have an xml parser that has an xml catalogue of some sort that recognises say XHTML or Docbokk PUBLIC identifier, the file will work even if the system ID points at a file not accessable to you as your parser can use the catalogue to map the public ID to a local copy. Conversely if you don't have that public id in your catalogue (or if there is no public id) then you can perhaps find the file directly by using the system id.

When the syntax was invented, for sgml, this last step would most likely fail as it would be a local file name so it would assume that you had local files in the same position. In XML when the SYSTEM is supposed to be a URI and so perhaps usable from anywhere in the world there is an argument (that I don't personally hold, but which is not unreasonable, and comes up often) that the SYSTEM ID is all you need and PUBLIC IDs no longer serve a purpose.

David

 

[Razvan]

Hi !

Look what I've found on a web site:
(http://www.adp-gmbh.ch/xml/dtd.html)

SYSTEM
<!DOCTYPE root-element SYSTEM "http://xyz.qq/abc/def.dtd">

SYSTEM means that a validating parser needs the dtd when it validates an XML document.

PUBLIC
<!DOCTYPE root-element PUBLIC "path/abc/def.dtd">

PUBLIC means that a validating parser might use a 'cached' ????
version of the DTD.

Mixing
<!DOCTYPE root-element PUBLIC "path/abc/def.dtd"
"http://xyz.qq/abc/def.dtd">

This form uses http://xyz.qq.abc.def.dtd only if path/abc/def.dtd is not available.

Regards,
Razvan

 

[David Carlisle]

Look what I've found on a web site:

Don't believe everything you read

if in doubt go back to teh definition:

http://www.w3.org/TR/2004/REC-xml11-20040204/#NT-ExternalID

[75] ExternalID ::= 'SYSTEM' S SystemLiteral
| 'PUBLIC' S PubidLiteral S SystemLiteral

which says that an ExternalID is either SYSTEM followed by a literal or PUBLIC followed by two literals separated by white space. there is not an option of having PUBLIC followed by just one literal. (except for the declarations of notations which also include a special PUBLIC-only production 83).

David