pcplantdb

[Chad Knepp <pyg@galatea.org>]

Bear K writes:
 > Hi everyone,
 > 
 > Thanks for all the good and interesting emails today. Do we have an 
 > archive up that Rich can use?
 > 
 > News on the personal front, I took a bit of a fall today on my bike, 
 > got a bit scraped up, and my hand is sprained, so I won't be typing as 
 > much (or at least as fast) for a little while. But don't worry, I'm 
 > still here. :-)

Ouch, take care.

 > Good to hear of your XML-RPC escapades Chad. I think this will come in 
 > real handy when we want to build out a TouchGraph client. I'll need to 
 > luck at the engine a bit more, but I believe data comes via XML-RPC.

That's the line of thought I was having.

 > Regarding Andy's project, we should research what XML/RDF formats have 
 > been developed for Event/Orginization information, and perhaps think of 
 > using a standard for interoperability. I know there's work in this 
 > arena going on.
 > 
 > Regarding book references, etc.....It seems like the text entries 
 > should perhaps be considered roughly equivalent to comments. One thing 
 > Jim talked about was reputation and knowing the source of info. Which 
 > is what the footnotes provide. We know the source, and who entered it. 
 > At one of the PlaNetwork conferences there was discussion of Douglas 
 > Engelbart's (a net pioneer) vision of the web was/should be back in the 
 > 70's. Thinking that each paragraph, or sentence of text should be 
 > referencable via URL, so people could comment on, quote, etc it. 
 > Perhaps this is something we could consider? Sentences come from a 
 > specific book, entered by a specific person, referring to Plant Habits. 
 > People could append a comment on that, say they've experienced 
 > something different, etc. Clients/users could opt to view just the 
 > upper level/high "informative" score information. How hard would that 
 > be to implement such a fine grained hierarchy in an RDBMS?

The concept of being able to attach a comment to every/any 'object' is
a concept I am quite into.  Defining object granularity as a sentence
is a little smaller than I had in mind.  I was thinking more along the
lines of a comment being the smallest object, but this could be
multiple paragraphs.  

While I agree that knowing the source of information is important, I
personally have more faith in the reports of personal experience by
Hemenway, Holmgren, Pittman, London, etc. than a reference to a book.
These folks already have reputations.  I also suspect that from this
point on most/all of the information we will accumulate will not be
researched from books like Ken Fern did with PFAF.  I've said all this
before...

Just checked and there are 270 books already listed in the dataset.
There are currently no references to anything other than a book.

This raises another interesting point, being what is the 'reference'
for the 30 - 40 little bits of data we have like height, preferred
moisture level, etc.  Currently they are the product of the research
and experience of Ken Fern, but how do we indicate that and how do we
indicate a change of one element of the thirty plus elements.  Some
things are missing and some things are, in my experience/locale,
wrong, so changes need to be made.  How do we track
accountability/reputation in this?

 > Also, on the "TouchGraph, grokking how to visualize this stuff" 
 > subject...
 > I hope people have kind of played with TouchGraph or systems like it to 
 > understand what I'm saying, but:
 > A user could "anchor" (so the nodes aren't pushed/pulled around) a few 
 > query items: "height is around 40 feet", "is edible", "nectar 
 > producing", etc. Place each in a separate part of the screen. Results 
 > would come in, matching at least 2 (1,3,controllable) properties. Items 
 > nearest abs(result height-40 feet)=zero are pulled to the height 
 > element with more force, also pulled to the other query items each 
 > result matches.  Thus results get clustered. The effect is like having 
 > a multi-axis graph, but you can use as many as you want. I think using 
 > something like this as the basis of the UI is nice. It's simple enough 
 > to understand, implement, and is relatively powerful. You could also 
 > anchor a few species (/elements) down, and see the relationships and 
 > possible connections between them. How to quickly select from available 
 > query options is a little UI question.

I grok, and I don't think this type of use/interface is limited to GBI
stuff.  The problem with text interfaces is that it's inherently
hierarchical, you have to put something at the top and follow it by
something.  The visual abstraction created by GBIs allows the for the
perception of equality or fine gradations (and groupings) of
hierarchy.

Regarding your last sentence: heck yeah!  There are probably 100
unique such options already in Vocabulary.py.  One possibility that
may or may not appeal to you is to keep the query text based.  Doing
this would require additional development and documentation of the
vocabulary in addition to an improved grammer but would really improve
the tool all around.  Think sub-language... something like 'plant:
height>30 and height<50' for your above example.  

 > OK, enough typing, need to rest my hand.
 > 
 > Cheers everyone,
 > Bear

-- 
Chad Knepp
python -c 'import base64;print base64.decodestring("cHlnQGdhbGF0ZWEub3Jn")'