[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: suggestions

Ciao

> Any thoughts on how to make the mailing list less quiet?

I hope the Tellable thing together with CVS release should
wake up developers (if there are any :-).

> There will ultimately be cases where two nodes represent the same
> resource, but one has no (or partial) URI associated with it in the
> data. But their identity can be inferred from context, eg. they're both
> onlyChildOf some other node. I guess applcations would just rewrite
> their URI using setURI and equals() would pick up on equality from that?

Sounds like heuristics to be applied so the application level 
interaction through setURI sounds reasonable.

> Another topic: remind me what SiRPAC's current behaviour w.r.t. relative
> URI resolution is?

The online version does indeed blur the current behaviour but
if you run SiRPAC from command line or in embedded environments
and you remember setSource() before running the parse() method,
SiRPAC will append all local names to that base URI.

- generated ids a la genid1, genid2 will become basenamegenid1
  basenamegenid2 etc.
- local namespace declaration xmlns="" will assign local IDs to
  this basename
- local description about="" will assign the basename to the
  attribute value

It is, however, not so advanced that if you said

about="../../foo" that it would resolve that part. Perhaps
I should implement that, too. I wonder if the Java URL class would
do that for me? Have to check.

> A Tellable being informed via the anonymous assert() interface wouldn't
> know the URL of the context document being parsed by the DataSource, so
> couldn't do this itself. (For round-tripping, you might want to also
> hang on to the relative form of these URLs somewhere too...?)

I agree, SiRPAC needs to do this right.

> Does SiRPAC have the notion of an anonymous resource? Or one with a
> generated (private) ID? This would I think be really useful, eg. for
> writing out again. Currently we can't tell 'urn:importantname' from
> 'file:genid_4' mechanically. (I use 'file:' prefix for generated IDs so
> they're legally URI)

All non-user defined nodes will have a system-assigned unique
id within that document scope (and eventually in the URI information
space once they are appended into the basename as I described earlier).

> Finally, is there likely to be demand for assert(Triple t)? I'd say not,
> too bloaty, but it was a thought that occured...

Again, I would prefer one single method users need to implement.
I would feel unconfortable implementing interface methods I would not
need.

Janne
--
Janne Saarela           | Visiting scientist
Email: jsaarela@w3.org  | World Wide Web Consortium http://www.w3.org/