Subject: [TS] Musings from France
From: Ben Darnell (ben@thoughtstream.org)
Date: Wed Jun 28 2000 - 10:24:55 EDT
Here are some notes I made about ThoughtStream while I was in France.
It starts out with simple practical changes, then gets into more
philosophical design issues. In short: I'm starting to come around to
the Topic Map way of doing things (provided I can overcome some UI
hurdles) (Jeff, you might want to hold off on those ZClasses until the
dust settles here)
--------------------
* Inline links (wiki-style?)
* Scrollbars on the note edit screen
* Adjust borders on note edit screen
* Crash when changing a label
seems to work, but creates an extra item.
* Use multipart/related for the contents of an Idea? How does this
interact with xml?
* Fine-grained links
This could be done with XPtr or mime content-id. Alternately, promote
content objects to 1st-class objects and use xlink:show="inline" to
link to them (as per Andrius' suggestion). Thip would make anything a
potential link target, at the expense of a lot of extra bookkeeping.
* It's tedious to create all the date ideas by hand. I wonder what
options there are to automate that?
* Now I really see what Andrius was getting at with the
sequence/hierarchy/network business. I'm trying to do a sequential
narrative with a navigational network overlay, and it's not working
terribly well. This is partially due to the coarseness of the linking
capabilities that exist at this time, but I wonder if there might be
deeper issues here.
All this could be solved if I just broke things down into smaller ideas,
but that takes a lot of extra work. Specifically, every idea now
requires a title, which must be entered by hand. Perhaps the answer is
to allow untitled ideas, then all current Content objects become Ideas
of their own.
This breaks the idea behind the current browse model, which can find any
Idea by title. Maybe the titular browse window is the wrong way to go
about things. The main view is already a browser for all practical
purposes; why limit the 'browse' function to ideas with known names?
There must of course be a find-idea-by-title function, but I think it is
better to look at it as a search method rather than a primary means of
navigation.
An Idea then consists of:
* bookkeeping data - ID, timestamps, etc
* either one content object or a title. A title is like a content of
type text/plain (does it make sense to allow other data types here?
Topic maps do; it may allow some interesting applications, but
increases the burden on the application. Also, since titles are used
for locating ideas, it doesn't sound very useful to have an untypeable
title), but its presence signals that the idea is special in some sense.
All ideas must have either a title or point to an idea with a title (is
this necessary? It makes things rather like the curent model (at least
from a UI perspective), and I think without this restriction we risk
unnecessarily complicating the UI. Does this restriction prevent any
useful construct?). Further questions: does it make sense for an idea
to have both a title and another content object? I don't think so. How
about neither (such an idea would serve as a hub for links)? This could
be useful, but I think the same effect could be achieved with an Idea
with an empty text/plain content.
* Zero or more links (the zero-link case is probably useless, but can't
be disallowed, since at the beginning there will be only one Idea in the
mind) These links are esentially the same as the curent
model.
* Pressing D on an link with
different link labels does not put the inverse label in the
LinkLabelField
* In an earlier note, "Now I really see...", I see now that what I was
suggesting is essentially the same as (a subset of) Topic Maps, with a
slight extension to allow occurences to be specified within the same
document (not strictly necessary, but I think it would be a good idea.
Perhaps a real extension is not needed; a combination of standard XTM
with mime multipart/related).
This returns to mind the prospect of turning TS into a topic map editor.
Can this be done, or are there insurmountable hurdles? Most such
hurdles would be located on the Palm side - but the functionality of the
Palm is by the very nature of the device a subset of that of the PC, so
it is understandable if some of the more esoteric features (like
non-text/plain titles) are excluded from the handheld. What do topic
maps have that TS doesn't?
* complex titles
* 'facets' (I still don't understand what a facet of a topic is)
* much more potentially complicated links
* scopes
From a UI pespective, all of these things seem rather complicated. I
wouldn't want to do something like that unless it provided substantial
benefit to the user. I'm not sure that any of these capabiities
do.
-Ben
-- Ben Darnell ben@thoughtstream.org http://thoughtstream.org Finger bgdarnel@debian.org for PGP/GPG key 1024D/1F06E509------------------------------------------------------------------------ Challenged with e-Business quality management? Register for a free Webinar featuring e-business testing and performance experts. http://click.egroups.com/1/5935/7/_/6321/_/962202298/ ------------------------------------------------------------------------
This archive was generated by hypermail 2b25 : Wed Aug 30 2000 - 22:01:01 EDT