I think it's a good initiative.
El 24/05/18 a les 13:14, Camille ha escrit: > Hi all, > > Still on the topic of file format. I think it could be a good idea to > have a validating dtd for kdenlive: that could help diagnosing > wrongly-formed files and would also be a kind-of-human-readable > documentation for the format: as I couldn't find any in the source tree, > I started making one from mlt-xml.dtd. Do you think it would actually be > useful ? > > It seems it shouldn't be too much work. I tried to validate a simple > Kdenlive file against the mlt-xml.dtd and "fixed" the errors of the fly > ; I found two categories of issues: > > - Adding attributes and children elements, which is straightforward. I > always assumed CDATA for the type (so there is room for optimisation if > we want stricter control: e.g the hide element could have a more > constrained type, I guess). Just a few points that should be addressed : > - In the DTD, Playlist can't be empty, as they must have a least > one of the following elements : (entry | blank | producer | playlist | > property | tractor | multitrack)+. But I get an empty one in my simple > example : which is correct ? > - In the DTD, tractor has a mandatory multitrack child element : > tractor (multitrack, (filter | property | track| transition)*) , but > such child element never appears in my example. Should it be tractor > (multitrack | filter | property | track| transition)+ ? > > - ID /IDREF types issues. It seems that most issues related IDs has been > fixed between "Generation 2" and "Generation 3" formats. There is one > remaining, "main bin" as you can't have spaces in IDs > (https://www.w3.org/TR/REC-xml/#NT-Name ). could that be fixed in the > code, or should we give up on the ID/IDREFS types and just use CDATA ? > > Cheers, > Camille > >
