Author: | Benja Fallenstein |
---|---|
Date: | 2002-10-31 |
Last-Modified: | 2003-03-31 |
Revision: | 1.1 |
Status: | Accepted |
We need fake spans to store "computer data" like numbers representing cursor positions or window coordinates. This PEG presents a way to do fake spans in a way that is efficient, yet allows them to be xanalogically linked and transcluded.
Fake spans should:
Create "miniblocks" whose whole content is contained in their ID. In the file format, a fake span currently could look like this:
[FakeTextSpan, "214"]
A real span looks like this:
[TextSpan, "storm:block:01E88CEE7CF19F016EEF00B315C0B930C953DB7EF2", 0, 3]
Under this proposal, the fake span could look like this:
[TextSpan, "storm:inline:uzPa+vB7kMBWMZWw-Og58D2NXGeZ:1::214", 0, 3]
That is, in the block's ID, we give:
The syntax of the URI would be like this:
storm:inline:<RANDOM-BYTES>:<COUNTER>:<MIME-TYPE>:<CONTENT>
where <MIME-TYPE> may be blank in the case of text/plain, and <COUNTER> has the same role as in a urn-5 and may be omitted. <RANDOM-BYTES> must be encoded in the same way as in a urn-5.
Since all fake spans then have xanalogical ids (the URI of a Storm block plus the offsets), links and transclusion resolving works perfectly fine.
Whereever the id of an inline block (colloquial: "miniblock") goes, the whole content of the block goes. This means that when publishing a block that has a reference to a miniblock, the miniblock also gets published. Also, when transcluding a single character from a miniblock, the whole miniblock is always carried along.
To alleviate these problems, it should be respected that miniblocks must:
- Benja