Users might want to ask not just "what are all the checker pipelines installed", but "what languages have the punctuation pipelines" or "what languages have plain analysis pipelines".
But with a free-form attribute, pipespec writers might put in type="Punctutaion" or type="Grammar Checker" which ≠ type="Grammar checker" ≠ type="Grammar checking" etc. Not very queryable. OTOH, it might be nice to have some freedom sometimes – we don't want to constrain pipespec writers from inventing completely new things (type="phplint" using only CG? type="Cheek-movement-to-text"?).
Maybe we could let type be a closed set, and then the type Other requires an attribute for the freeform field like other-type="Blockchain parser"
Users might want to ask not just "what are all the checker pipelines installed", but "what languages have the punctuation pipelines" or "what languages have plain analysis pipelines".
Currently, the dtd gives a
type
on thepipeline
element: https://github.com/divvun/libdivvun/blob/master/src/pipespec.dtd#L42 but it's free-form and we don't actually use it yet.But with a free-form attribute, pipespec writers might put in
type="Punctutaion"
ortype="Grammar Checker"
which ≠type="Grammar checker"
≠type="Grammar checking"
etc. Not very queryable. OTOH, it might be nice to have some freedom sometimes – we don't want to constrain pipespec writers from inventing completely new things (type="phplint"
using only CG?type="Cheek-movement-to-text"
?).Maybe we could let
type
be a closed set, and then the typeOther
requires an attribute for the freeform field likeother-type="Blockchain parser"