[sc34wg3] Comments on N0975

[Lars Marius Garshol]

This is my comments on the proposals from the discussion in Kyoto, as  
summarized inhttp://www.itscj.ipsj.or.jp/sc34/open/0975.htm.

In general, this document contains little new that we did not see  
raised in pre-Kyoto discussions. So there will not be that much new in  
my comments, either, I'm afraid.

In general, I think avoiding changes is better than making changes.  
I'm quite happy with CTM as defined in N0935 and see no need to change  
it. In addition, the less we change, the smaller the chances are of  
causing syntactical upsets in TMQL and in CTM itself. So overall I  
favour a conservative approach.

--- Delimiters for topic blocks

I agree that the choice between two newlines or period is not good,  
and that two newlines (as pointed out by Xuan as well) is tricky to  
parse.

The C-style proposal of curly braces I find downright ugly. The period  
was much nicer, and it was precisely that it was not too visually  
distinctive that I liked about it. The period gives users more  
latitude to format their CTM as they will to emphasize the structure  
of their topic map.

The curly braces are used in TMQL to signify transitions between XML  
content and TM content. This is a convention borrowed from XQuery  
together with the FLWR, and personally I consider this one of the  
strong points of the current TMQL syntax. Using the same characters  
for blocks in CTM is confusing, and something I feel very strongly we  
should not do.

--- Delimiters for statements

I see no need for this. I find the semicolon ugly, and would much  
rather we did without.

--- Lists of values

I don't mind this, as long as it doesn't cause any problems elsewhere.  
(Interaction with template invocations?)

--- Delimiting IRIs

I'll stay out of the delimiter issue for now.

I agree occurrence URIs should always be given as literals.

--- Binding subject identifiers to local identifiers

This is incomplete. "The WG agreed with the proposed syntax for  
binding subject identifiers to local identifiers and extended the  
proposal as follows: ..." What syntax proposal? What is "binding  
subject identifiers to local identifiers"? What, in short, is the base  
that is being extended? So most of this I can't comment on.

Whether we use ~ or = to mean subject identifier/locator I don't  
really care. Whatever.

Using '#' for item identifiers is, well ... Some quick comments:

(1) This is asking for trouble, given that these appear inside
     IRIs, too. (I'm assuming we ditch the <> wrapping.)

(2) This would make the *perfect* comment syntax. *Far* better
     than "!", which as far as I know hasn't been used for comments
     since Simula 67.

I won't attempt to say how much I dislike this proposal, but I would  
feel much happier if we didn't do this. Some other character, *please*.

--- Multiline comments

Why have them? We'd be better off without.

--- Reifier syntax

I don't care that much. It's not pretty, but I have no better proposal.

--- Template imports

This is fine with me. (In fact, I thought this was already accepted.)

"A file may not be included more than once in the same document."
  This doesn't need to be an error. In fact, it might be better to
  just accept it without complaint.

  That would make it possible to do the following:
    base.ctm (includes nothing)
    foo.ctm (includes base.ctm)
    bar.ctm (includes base.ctm)
    master.ctm (includes foo.ctm and bar.ctm)

--- Minor stuff

undef, meaningful whitespace: fine

Unicode escape sequences: what is the issue? I don't know of any  
problem with this.

--- Colons

I preferred the colons.

--- Template definitions

I preferred it the way it was.

--Lars M.