Re: [xtm-wg] A challenge on "the graph"

[Lars Marius Garshol <larsga@garshol.priv.no>]

* Lars Marius Garshol
|
| - it is far removed from the actual form of implementations, making
|   it much harder to understand for implementors, which are after all
|   the intended audience

* Sam Hunting
| 
| 1. Property sets are quite close to implementations, yet can be
| quite close to implementations, and describe graphs.

If you compare the two models with the APIs of tmproc, TM4J and jTME
you will find that the property-based model is far closer to their
APIs than is the graph-based model.

It is of course possible to implement topic maps using some API
designed like the graph model, but it would be highly unusual and also
very unlikely to be convenient for client code.
 
| 2. Implementors are NOT NOT NOT the only audience. What would be the
| use of a spec that could only be understood by implementors, but not
| the people who purchase and use the implementors products? 

I don't feel very strongly for the end-users as an audience, but even
if they were to be counted as part of the audience for the model I
think they should prefer the property-based model.

* Lars Marius Garshol
|
| - it is insufficiently detailed in that some properties of the nodes
|   in the graph are left out, even properties of great importance to
|   processing
 
* Sam Hunting
|
| Specifically?

Where do the contents of the <baseNameString> and <resourceData>
elements go? Where do the URIs of each node go? Also, where does the
URI of <resourceRef> go?
 
* Lars Marius Garshol
|
| - the form the graph is specified in (property-less nodes
|   connected by arcs with endpoints of two types) causes the
|   description to become difficult to follow, since everything is so
|   type-less and so very similar everywhere
 
* Sam Hunting
|
| Not so! Before making such indictments, please STUDY the existing
| processing models -- the DC version and my straw man (which I assume
| Kal gave you?). 

He did. I've studied it, but failed to comprehend it.

| There are types for nodes, and types for arcs. 

I know there is, but everything is composed of nodes and typed arcs,
and I found that difficult to follow, both in SRN's version and much
more so in your very compressed version.

| I think "very similar everywhere" is an aesthetic judgment that will
| dissipate with study. 

It is the impression it gives. Of course there are differences, but
they require an effort to see. I feel convinced that that effort is
not necessary in a property-based model.

| With everything concentrated in one place, the differences leap into
| sharp relief.
 
The only thing that leaped out when I looked at those tables was
confusion. The fault may be mine, but that was the effect.

What I would like to see, Sam, is your arguments for why the
graph-based approach is better than the property-based one, if that is
indeed your opinion. And what I would _really_ like to see is the
opinions of the other members of this list.

* Lars Marius Garshol
|
| This vendor is not Ontopia. It may be some other vendor, but if so I
| know nothing whatever of it.
 
* Sam Hunting
|
| Actually, this worries me more than ever. Here's the thing -- I'm
| making a real effort to understand everyone's ideas -- surrenduring
| one's fixed views builds character and an open mind -- but I don't
| sense an equivalent effort being made everywhere. (NOT you
| personally, Lars -- we just had some agreement on requirements. ;-)
| 
| THIS IS IMPORTANT STUFF! All ideas should be looked at, even the ones
| we don't like or find ugly.

Who are you speaking to here? Not Ontopia, I hope, since I have
already told you that we do not have such an undisclosed review.

--Lars M.


------------------------ Yahoo! Groups Sponsor ---------------------~-~>
Do you have 128-bit SSL encryption server security?
Get VeriSign's FREE Guide, "Securing Your
Web Site for Business." Get it now!
http://us.click.yahoo.com/2cW4jC/c.WCAA/bT0EAA/2n6YlB/TM
---------------------------------------------------------------------_->

To Post a message, send it to:   xtm-wg@eGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@eGroups.com 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/