RE: Reporting Errors from Extensions [and xalan functional-programming]

[Joseph Kesselman]

>So may be this is a good chance to ask you Xalan guys what your view on
>the functional aspects of xslt are? Do you think that the functional
>aspects of xslt are unimportant?

Personal opinion follows, which may not match that of any of the other Xalan developers:


On the contrary; I consider XSLT's functional nature very important. We're probably going to have to take advantage of that as we get more serious about stylesheet optimization. Folks should be assuming that XSLT stylesheets may be executed "out of order", because someday they probably will be.

Yes, the SQL extension currently cheats on that rule. Its use of side effects allows a lower-overhead solution than would otherwise be possible, at the expense of relying to some degree upon (a) an understanding of Xalan's current behavior and (b) restrictions upon how this extension can be successfully used. Nothing wrong with that as an engineering solution, but it does mean that we may have to rewrite it as Xalan's behavior and its users' needs change.

Personally, I'm having trouble thinking of any situation where I would _WANT_ my stylesheet to be altering the filesystem -- except by creating new files (via main output or the redirect extension), which may or may not imply creating the directories those files will occupy. . XSLT's purpose is to transform input documents into output documents; if you're doing something else (such as reorganizing your filesystem), you probably should be doing that in a different language either before or after your XSLT stylesheet runs, (Perhaps having the stylesheet generate a document which describes the actions to be performed, and invoking a separate processor which runs those actions.)

______________________________________
Joe Kesselman / IBM Research