>This is just a set of current cramps floating around in my brain:
- It is still amazing how Validation with Models is still misunderstood. It doesn’t matter what type of model, the common conception is that it slows down performance so model validation is not done. Unfortunately, with out validation of models before processing them, you can allow bugs to creep down further into the system than if you had validated to begin with. When I’m speaking of models I’m speaking of any type of model and modeling framework. Whether that be EMF, DSL, XSD, RelaxNG, DTD, or Schematron.
- Many projects at eclipse will run into some problems with their builds when they make the migration to git. The concepts and assumptions that have become ingrained with CVS and SVN will not necessarily work with the way git works. An example is the way the WTP source editing structure is currently setup? If you want to do component based builds on say the wst.xsl components, you will run into some problems. More on this in a future blog post.
- In the coming months I see concepts and build systems that try to remain agnostic to the way the files are stored in the IDE and repositories, having an advantage over those that rely on specific layouts. To this end I see a good future for Maven3/Tycho, as well as B3 if it keeps from being too eclipse IDE centric.
- Most programmers do not realize that the SAX EntityResolvers will not kick in unless there is either a DOCTYPE declaration or schemaLocation attribute specified. This is especially bad when trying to do validation of an XML file against a set of grammars. Most XML will not come in with a schemaLocation attribute so even if you put the XML Schemas in an EntityResolver you will not get validation to occur. The common error will be a message during validation that it can’t find the root element. There are better ways to work with a known set of schemas, and to avoid performance pentalties. Grammar Pools. More on this in a future post.
That’s about it for now. Back to the grind.