> Depends on whether you are describing the whole file or the
> individual declarations in the file. The whole file *is* a parameter
True, but so is (for example) tei2.dtd when referenced from
> I didn't name them, I just put them on a server. I'm sure Rick had a
> reason for .pen. Perhaps he uses .ent for general entities (where
> others might use .xml, making it harder to tell non-document
> entities from document entities).
Sorry, I wasn't trying to apportion blame, I just felt that a filetype
which described the content of the file ([character] entity
declarations) might be more appropriate than one which described how
you got there. ISOlat1.charents would be more like it.
> A good argument for FPIs, if you ask me :-)
Yep, except that at some stage it has to be dereferenced to a local
filename. I rather liked the old sgmls algorithm by which
-//Davenport//DTD DocBook V3.0//EN
but most OSs have problems with filenames containing [/:()]+ and would
definitely gag when the WebSGML Adaptations are applied to raise the
bar on minimum literals for FPIs...