A conceptual question: if a user provides an image query that can't be exactly matched to an image, it is left as is; as text within img tags. This means that the Hub can produce ScML that is invalid according to our DTD, which I think we should probably avoid (in theory, any ScML file we create should be able to go to any of our output streams without validation problems).
I see four options:
1) Allow this particular instance of invalidity.
2) Allow non-empty img in the DTD.
3) Present this different in the ScML output; for instance, remove the text from the img tags and make it a child of fig, or make it a comment.
4) Raise an error and put the burden on the user to correct their input.
I'm not sure which I prefer (I lean towards 3 or 4).
A conceptual question: if a user provides an image query that can't be exactly matched to an image, it is left as is; as text within img tags. This means that the Hub can produce ScML that is invalid according to our DTD, which I think we should probably avoid (in theory, any ScML file we create should be able to go to any of our output streams without validation problems).
I see four options: 1) Allow this particular instance of invalidity. 2) Allow non-empty img in the DTD. 3) Present this different in the ScML output; for instance, remove the text from the img tags and make it a child of fig, or make it a comment. 4) Raise an error and put the burden on the user to correct their input.
I'm not sure which I prefer (I lean towards 3 or 4).