[topicmapmail] [ANN] Topic Map Explorer Version 1.0

[Thomas B. Passin]

Jason.Markos@empolis.com wrote:

> This mail 

[Jason refers to my previous post about the Topic Map Explorer product]

> seems to be at odds with itself.  On the one had it discusses the
> interest and validity in the idea of having a personal desktop application
> whilst on the other hand, it criticises the applications lack of end user
> support for TopicMap specific constructs such as roles, scope and typing.

That is not inconsistent.  You can have a good idea that does not live 
up to its potential (or has not reached it yet).

> In my limited experience of working with TopicMaps the real challenge has
> been how to make it accessible and useable by average computer users - which
> this particular application seems to be aimed at.  (The documentation
> provides an interpretation of TopicMaps making it useful and accessible.)

It is definitely a challenge.  That's why I wrote TM4JScript and the 
applications built on it, so as to have a vehicle for exploring how to 
make TM maps and apps usable.

> In which case, surely the question should be - does it help users in their
> day to day life?  It would appear that the application may not have specific
> TopicMap constructs visible to the end user 

I do not claim that all topic map constructs should always be exposed to 
  users all the time.  Not at all.  But if when I create a relationship, 
it has no roles and can only be binary, then I could have done the same 
thing with rdf (as one possibility).  In fact, so far as the exposed 
interface is concerned, all one has is a generic semantic network. 
That's not necessarily bad, it just doesn't use topic map specific 
features or concepts.

Now Rick Coope has said that under the covers, the thing is built on iso 
13250 - or maybe it is built on the data model that may or may not 
become part of iso 13250 - or something like that, anyway.  Well, that's 
good, and maybe more 13250 concepts will get exposed over time.  That 
will be very good, too.

> - but it certainly seems to
> allow me to define topics, relationships, link them to content 

I will allow as how the feature of bringing in files or file trees is 
pretty nice.  But even here, you have to accept the one relationship 
type that the program thinks is the right one.  I'm not sure how useful 
that all is, though I am willing to spend some time here to get more of 
a sense for it.

> and to build
> network based structures as opposed to linear/hierarchical structures - all
> TopicMap principles.  Without requiring me to be a TopicMap expert - which
> many of you can confirm I am not!
> 
> If XTM is to become pervasive as a standard, it will need applications such
> as TMExplorer that make it accessible.  You only have to look at HyTime to
> see what happens to standards that are inaccessible.

You will get no argument from me on these points.  Nevertheless, I stand 
by what I said in my original post.

For example, Rick said that the app has a notion of topic identity.  How 
is this manifested and how can I make use of it?  Not by name, I mean by 
referring to some other topic or resource outside of the topic map 
itself (whether psi or whatever)?  Nope, can't be done, not in this 
incarnation.

Well, OK, this is a stand alone desktop personal app, and most of the 
toy or small maps a lot of us make to experiment with don't refer to 
anything outside themselves, so for a stand alone app, maybe this is OK 
for the time being.

Well and good, I suppose, but you can't claim that this incarnation of 
the map exposes anything that bears on topic identity except for a name. 
  Is it present under the hood? Does it make any difference?

As for the app not being production quality, let me give a couple of 
examples.

I create two topics - Huricane and Weather Event.  Now I want to relate 
Hurricane to Weather Event.  I have already created a relationship type 
named "a-kind-of".  I click on "Start" in the left-hand panel, so as to 
display all topics in the right pane (all two of them in this case).

I right-click on Hurricane and from the popup menu I select "Set 
Relation".  This gives me another menu with items that include 
"a-kind-of".  Just as you would think it might work.  I select my 
relationship type.  Boom! I get an unhandled exception.

Luckily, the app allows me to choose to continue.  Should I do so or 
not?  Hmm, I will probably decide to go ahead.  Is that a good idea or 
not?  I can't tell.

Now I realize that the sequence I described above is not quite what the 
user guide describes.  But I should be able to select those menu items, 
and I should not get an unhandled exception for doing something it is 
obvious most users are going to try sometime soon.

In point of fact, to create a relationship I have to create one with the 
default type (using a different method than the above sequence) and then 
change it later to the type I want.  Not intuitive.

Now I want to set the type of a topic (as opposed to using my a-kind-of 
relationship).  I right click on the topic and get a menu item saying to 
set the topic type.  I click on it, expecting to get a dialog letting me 
select the type.  But instead, nothing happens.  At least there is no 
exception this time.

Hmm, according to the user guide, I have to create an "index" with the 
type I want, and assign that index entry to get a type.  Huh?  In topic 
maps, a type is a topic, not an index.  What is an "index" in this 
context?  After reading the guide, I am still unsure about how an index 
relates to a topic relates to a topic type.

So again we have the same pattern - menus items that seem to allow for 
the obvious action but then do not work, while the right way is non 
intuitive and (in this case, at least) apparently not in accordance with 
how topic maps work.

Yes, under the surface the implemenation could be fully topic map 
compliant (let's not argue here about exactly what that means!), but not 
so far as the UI, the user experience, is concerned.

Now if the app could export and import xtm topic maps, at least it could 
claim to be more or less interoperable on an xtm level.  That would 
certainly be worth something, quite a lot in fact.  But alas, not in 
this incarnation.

So I repeat, this is not production grade software, and not "topic map" 
software.  Not yet.  And please notice that I am so interested in having 
good, affordable examples of this kind of software that I got it and 
installed it within minutes of reading the announcment.  If I am harsh, 
  it is because I am so interested in someone getting it right.

So I would encourage the developers to suspend downloads and sales until

1) They get rid of the unhandled exceptions for perfectly normal 
actions, like using enabled menu items, or closing the app (I get an 
exception on closing more often than not).

2) They disable menu items when they are not appropriate or will do nothing.

3) They provide export/import of xtm.  I don't say they have to 
implement every feature of xtm, but at least the app should be able to 
extract the parts that it uses - topics, names, associations, and 
occurrences.

4) They document their rdf export/import format, if they continue to use 
that as well.

After this, then let them make it available again, and we can all see 
what people think about it after some use.

Cheers,

Tom P

-- 
Thomas B. Passin
Explorer's Guide to the Semantic Web (Manning Books)
http://www.manning.com/catalog/view.php?book=passin