Hi,
Good idea! Notice that test failure is not the only reason to set an lpe as experimental however: the definition of parameters and their internal representation are also relevant.
With this respect, "hatches" are clearly experimental atm, as the parameter list is very bad...
The tricky case is "knot".
it has 2 parameters: the size of the "gaps" and an array storing the "sign" of each crossing (the user can choose which strand is above/below the other, independently for each crossing).
The first one is typically global: we can have several knots in the same drawing, and want to have similar "gaps".
The second however is typically local: this information only makes sens for one particular path. sharing it among different paths is meaningless.
That is what makes me think lpe should be allowed to register "local" parameters. A new object attribute could be defined (something like "lpe-local-params"), where lpe would register all the data they need.
Moreover, I think we'll need this sooner or later to fix the transform issue (lpe cannot be shared atm because the params are transformed along with the object; they'd better stay untransformed, and a local offset transform be stored in each object the effect is applied to...)
So I submit this question to our gurus: is allowing lpe to register "local" parameters a good idea? if yes, is it hard (I'have no idea how to do that)?
Btw, if storing local offset transforms instead of transforming parameters is the good way to go, I think I could work on it.
Hm... as for "knot" being experimental or not... I dunno; I'd really like to have it added to the next release (it earned inkscape some new users in the little world of 'low dimensional topology').
Cheers, jfb.