[Taxacom] New Interactive Key to Wetland Monocots

[Stinger]

Well I wrote a detailed response to  Dallwitz  yesterday but
it doesn't seem to have made it here, probably for the best. There is
no need for another long thread where we just don't agree.  I do
agree that a "Best" characters algorithm is needed in SLIKS and it is
on the todo list. There is a significant tradeoff in performance though
and only  Dallwitz and now Watson have "asked" for it
specifically.  The assertion of Dallwitz that it is impossible in
a binary data structure is just wrong. It is actually much easier to
optimize character  states because they are the base unit. There
are no best characters in the real world, only best character
states.  

The assertions that SLIKS is only capable of using binary characters
and giving one sided choices for binary characters applies only to the
version released with the wetland monocots. Other versions 
available for free at the website  can have up to 9999 
character states for every character.  Binary data was used in the
wetland monocot data because it is an order of magnitude easier to
manage and curate on the back end, especially when numerous data sets
need to have overlap and algorithms for keying any possible combination
of taxa from different data sets have to be used. A more realistic
example in front end keys would be that if a character has 11
states,  in a character optimization paradigm, you would always
have all 11 states to choose from while in a character state
optimization paradigm, you would only have those that applied to the
remaining taxa.  You always have to choose among character states,
no matter how you optimize. LucID has gone much farther than IntKey in
recognizing that and if you're really  bothered by the issue I
suggest that you  transfer the data to LucID. It isn't difficult,
and there is nothing stopping you from doing it. 

That brings up the real issue here. SLIKS has never enjoyed paid staff
time or marketing  of any kind and no one has ever paid a penny
for any part of it. Yet the community is beginning to pick it up and
improve it. Not because of ignorance. We all know about many
keying  programs. I've distributed keys in 9 different keying
evironments including several DeLTa based ones. As in herbarium
databases, it is about the data not the program. The ignorance is in
doggedly sticking to a single paradigm in the face of change. 
Chris Meacham has a version of SLIKS and Greg Alexander wrote an
entirly new implementation called SAIKS that I encourage you to look
at. He used a data set for the prototype btw that has been through at
least 5 keying environments over the last 37 years. As it says on the
SLIKS web site, Intkey and LucID, especially LucID, are much more
functional and mature but neither meets my basic criteria for a program
that will have broad community acceptance. 

1. Free - all parts. Any charge at all keeps it out of the majority of
hands. It isn't the price, it's the fact that you have to buy it. 

2. Open source  (in the broad sense) so that the community can
make changes at will, creativity is encouraged, and a sustainable model
for longevity is built through community ownership.  The data
files should be open as well.

3. Completely web enabled with absolutely no plugins or installation
needed - it has to work like any other web page in your browser. 
Any impediment to the click-and-use (e.g. a plugin) reduces usage by
80%.   Many government and corporate computers are barred from
installation of any plugins - even the JRE. 

4. Extreme portability. It has to run on anything. 

4a. This is provisional because I keep hoping something better will
emerge but it almost has to be in javascript (not Java) for portability
reasons and ease of modification. Every modern browser (even those on
cell phones) supports javascript out of the box (although there are
some differences in implementation). Java is much more powerful but
won't initialize if conditions aren't right and has never really lived
up to its goal of running everywhere.  Unfortunately, many
advanced options (e.g. continuous character support and the best best
algorithms have a large CPU overhead in it so  performance
suffers.  It is a problem that hasn't been worked out yet. 

SLIKS is free and open.  If you want some other feature, you are
very welcome and strongly encouraged to make changes, many have, and I
am quite willing to help.  If someone wants to put in a grant
proposal to make a program that really meets ALL of the above criteria
and has all of the versatility to allow people like Mike to be happy as
well as everyone else. I'll help you write it and you can have the
credit and the cash as long as you deliver.