On Oct 28, 2008, at 11:00 AM, Sergiu Dumitriu wrote:
Vincent Massol wrote:
ok so far we have:
* Jerome: iii) * Anca: iii) * ThomasE (non-binding): iii) * Vincent: iii)
Sergiu, ThomasM, JV, Marius, Asiri (and other committers): are you ok with iii). Barring any more answers, I'll start implementing it in 1 hour (it seems the right solution to me anyway).
+1 for both ii) and iii)
iii) is good because it allows to have different URIs specify different actions. By the way, we could extend the URI prefix so that if there is no special handler for a prefix, we could use it as the action, for example: [[feed:Blog.BlogRSS]] would link to /xwiki/bin/feed/Blog/BlogRSS. At the moment this prevents FQN links to other wikis, so we should decide when to change the FQN syntax.
ii) is good because it allows us to specify custom attributes on any element/section, which would be needed for microformats (which rely on class, rel, and title mostly), styling, and with the use of namespaces we can add our own attributes and bind custom behavior on them. Anyway, this is not a priority now, just something for the (not so far) future.
I'd like to understand this better since I don't think ii) and iii) are perfectly compatible (this would mean two ways of settings parameters for links for ex which I don't like too much). So what's the exact need of microformats? How will it be represented? We can already do spans so that's a first solution and we can already put parameters on all our elements (lists, paragraphs, tables, links, etc). Then I'm also interested in implementing support for document level parameters too. We need a way to define the syntax for it so that it's different from parameters for a paragraph. For embedded documents it would be easy: (% param=value %) ((( ... ))) Last, I'm also interested in having the notion of properties as described in http://code.google.com/p/wikimodel/wiki/AdvancedStructuralElements . Maybe that could that help for defining microformats? WDYT? Thanks -Vincent
Implementation note: I'll still introduce an onImage() event and an ImageBlock class. Also as Jerome mentioned I'll use the text specified as the label as is.
Thanks -Vincent
On Oct 28, 2008, at 6:53 AM, Jerome Velociter wrote:
I'd go for iii), without parsing the label (use "hello **world**" if it is, not make it "hello world")
Jerome.
Vincent Massol wrote:
On Oct 27, 2008, at 2:01 PM, Vincent Massol wrote:
[snip]
Basically there are now 2 possibilities:
i) using a special syntax for parameters for links and images. For links that could be [[label>>reference>>>parameters]]. Note that we might also need such a special syntax for inline verbatim blocks. ii) using 2 generic syntaxes for parameters:
(% ... %) for applying parameters only to the next element {% ... %} ... {%%} for applying parameters to the elements inside
Which one do you prefer?
I think we have an issue for i) anyway (for verbatim inline elements for example).
My preference goes to ii) [EDITED: was i)]. After speaking to Mikhail (wikimodel creator) there's a third possibility:
iii) using the same syntax for links, images and attachments. Namely:
Links: [[label>>reference||params]]
Images: [[label>>image:wiki:Space.Page^my.png||params]]
Attachments (download link): [[label>>attach:wiki:Space.Page^my.png||params]]
This means that it would also be possible to use links, images and attachments directly in the text using the form:
This is a http://... URL This is an image:my.png image This is my word attach:my.doc document
Notes: * The "label" for images would correspond to the alt/title attribute. Thus the wiki syntax inside would be transformed into text (for example: "Hello **world**" would generate "Hello world") * It would currently not be possible to disambiguate between a subwiki named "image", "attach" and displaying an image or linking to an attachment. This would be fixed later on when we change the FQN (Fully Qualified Name) for Documents in order to support nested spaces. * This means that we can keep using (% %) notation and there's no need to define a new {% ... %} notation. * It means we need to introduce the "||" separator for link parameters * It is completely inline with the wikimodel way of handling images. * There's no need to introduce a new onImage() event. However we might want to rename our begin/endLink() events to beginReference()/ endReference().
I personally like this solution except for the fact that since we've introduced wiki syntax in place of labels we would be tweaking the model a bit since there's no need for begin/end for images. That said we could also add an onImage() event and emit it at the parser level so that the label would always contain pure text (if we have "hello **world**" we would use "hello **world**").
I'm hesitating between ii) and iii).
WDYT?
Thanks -Vincent
so +1 for 1)a) only if we can do it for all inline elements.
Which brings us to images: I find a wiki syntax quite suitable, for consistency reasons (with links, for example). ! is a pretty used character, even duplicated, so I'd go for another one: pipe (hard to find on some kbds, though), @ (semantically associated with emails, though), # (already taken), & ?
so +1 for wiki syntax for 2)
Happy coding, Anca
-- Sergiu Dumitriu http://purl.org/net/sergiu/ _______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs