On 16/04/2015 06:46, Bert Verhees wrote: > >>> >>> I think it is a stupid rule in the XML-Schema standard. >> >> I just hit this in doing the AOM2 schema. It's a completely senseless >> rule, clearly a hangover from 'document' thinking - nothing to do >> with 'data' thinking. I ended up replacing <sequence> with. >> >> <xs:choice maxOccurs="unbounded"> > > It is a poor man's choice, because of the side-effects. > The "choice"-constraint makes the engine ignore the > minOccurs/maxOccurs constraints in the elements under the choice element. > >> >> I have to say, XML-schema based data looks extremely unattractive as >> a basis for anything except data exchange. I wouldn't try to >> implement anything important inside a system with it, you are too >> compromised in too many ways. > > It is also used, as I wrote yesterday, to tell an XML-database how, in > a specific namespace, the data are arranged, in that way the database > can auto-create indexes, etc. > There is no other way to communicate the structure of XML-documents, > and even if we found another way, or invented it ourselves, we still > would need a broad acceptation.
We still need some kind of schema for XML docs/ messages, but I would not even think of making any persistence be based on XML, especially not XML schema. Probably Relax NG would have been the better one for messages etc. Keep XML at the boundaries, that's the key to happiness! I guess JSON with its own schema will replace it in the next couple of years. - thomas
