[topicmapmail] Web Services

[Murray Altheim]

Jan Algermissen wrote:
> Murray Altheim wrote:
> 
>>Jan Algermissen wrote:
>>
>>>Murray Altheim wrote:
>>>[...]
>>>
>>>>Do people have to be aware of the list of IDs, or does
>>>>the system just fail when either an expected ID isn't there, or
>>>>when an inserted ID fails because it's already in use (or there is
>>>>an error because of that ID).
>>>
>>>IMHO using element IDs for other referencing purposes than for reconstructing
>>>the serialized graph are a bad choice anyway (as are PSIs that contain fragment
>>>identifiers) due to the limitations they introduce when used over HTTP....
>>
>>Hmm. I'm confused about that last statement. If I publish a Topic
>>Map as a source of PSIs, or an XHTML document that references an
>>XTM document acting as that source, either way I have a single
>>XTM document containing multiple <topic> elements, where each is
>>potentially the Topic "behind" that the published PSI. I do this
>>because the PSI is yes, just a string, but the Topic is perhaps
>>firmly interwoven in a lattice of relations that might be important
>>to my processor. So those fragment IDs are important to me. Being
>>over HTTP or not doesn't seem to make a difference here. It's a
>>reference on a local file system too.
> 
> The problem with fragment identifiers is, that they are not part of the
> message when you invoke a HTTP method on them. The consequence is that
> no intermediary will see them and thus cannot add-in information it
> might have about that URI.

I don't mean to be snide, but that's not my problem. :-)

You are perhaps looking at this from the REST POV? I'm looking at
it from a server POV. The HTTP server certainly receives everything,
so the fragment identifiers do exactly what they should do upon being
received, i.e., they point at a specific ID within the resource. All
of this information is passed on to the application, right?

> Another problem is that creating flat namespaces (a single 'document' with
> lots of significant fragments) will allways cause the whole document to
> be transfered just to access the fragment (as said, the server will never
> see the fragment, the user agent strips it off).

Again, I'm looking at this from a server POV. The server doesn't have
to transfer everything if the ID is a part of a query.

> A more practical example is that you cannot, for example, submit a fragemnt
> identifier-URI to a search engine for indexeing.  foo#1 and foo#2 are
> exactly the same thing to them.

Same answer as above.

Perhaps we're talking about different things?

[I'll leave off discussion of encoding, since Lars Marius has
answered that already.]

Murray

......................................................................
Murray Altheim                    http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .

   US envoy warns of Taleban return, BBC News
   http://news.bbc.co.uk/1/hi/world/south_asia/3402813.stm

   When asked about concerns that the US call for "acceleration"
   [on building up the Afghani army to fight the resurgent Taleban]
   was linked to the timetable of American elections in November,
   Mr Taylor said: "We all remember what happened in the United
   States [on 11 September, 2001] and where those attacks came from."

   Saudi Arabia?  The attacks did not come from the Taleban, who
   are only concerned with control of their own country.