apologies and decision table

I apologize for upsetting John and a 
few others.

I suggest that the discussion has advanced
to a point where we can construct a decision
matrix that would simplify and take some or
all of the heat out of the discussion, leading
quickly to either a consensus or "consensus that
no consensus is forthcoming":

There are a handful of technical proposals
coming clear, that are candidates for a Recommendation.
These would be (each a Y/N question):

  1. same-origin + CORS?
  2. require ttf/otf?
  3. require EOTL?
  4. require WebOTF?

We can add or subtract to that list but those are
the most likely suspects.

They are 4 independent choices.  So, looking
just at those, there are 16 possible technical
recommendations.   The matrix has 16 columns.

        1             2             3          etc.
    not s.o+CORS    s.o.+CORS      s.o.+CORS
    not ttf/otf     not ttf/otf    ttf/otf
    not EOTL        not EOTL       not EOTL
    not WebOTF      not WebOTF     not WEBOTF


It's a little more difficult to label the 
rows but the rows are labeled in terms of 
a list of "proposed goals".

What is a goal?  It's a statement about the
practical impact on usage.  A list might include:

  a. obstruct casual download of restricted
     fonts from web to desktop?
  b. obstruct casual upload of restricted fonts
     from web to desktop?
  c. streamline casual download of libre fonts?
  d. streamline casual upload of libre fonts?
  e. require only trusted cross-site linking?
  f. convey licensing meta-data with font?
  g. encourage use compatible with unpatched IE<=8?
  h. encourage compatible use with other unpatched browsers?

We could again, add or subtract from that list.

Each box in the matrix is thus a goal intersected with
a possible combination of features.  For example, 
one box has the row label:

  "obstruct casual download?"

and the column:

  no on same-origin+cors, yes otf/ttf, no EOTL, no webOTF

The question "obstruct casual download?" can be answered
yes or no so we can fill in the box with two values, one
for each case.   

         no-case / yes-case
         ------------------
           good  /  bad

In other words, if we say "no, don't obstruct
casual download" then the case of no s.o.+CORS,
yes to otf/ttf, no to EOTL, no to WebOTF -- is good.
Perfectly acceptable.   If we say "yes, obstruct
casual download!" then that same combination of 
technical requirements is bad since it fails to do
the job.

For some of those boxes in the matrix, like the 
one I just described, the values to fill in ("good/bad"
in that case) appear to be completely uncontroversial.

Other boxes - not so uncontroversial.  The discussion
is not clear enough, to this point, for there to be
widespread agreement.

Some rows are not so very controversial.  If I'm
not mistaken, for example, nobody has any particular
problem saying "require only trusted cross-site linking? yes!"
That's handy because then we can just draw a big
black line through every column in which "s.o.+CORS"
has the value "no" -- those columns aren't worth considering.
Hey, look, we just cut the design space in half!
We only have 8 columns left to consider!

Some rows are not entirely independent of one another.
For example, "encourage use with unpatched IE<=8?"
is incompatible with "require only trusted cross-site
linking."  If we've crossed out all columns that
don't require only trusted cross-site linking, then
we can ignore the remaining "yes-case" values in
the row for "encourage ... IE<=8".   Instead of 
"good/bad" type values in that row, we can just
have "good" or "bad".   So we just trimmed the 
design space further.   In fact, we can do similar trimming
for the row "encourage unpatched use of other browsers".

Some rows in the matrix are controversial.
For example, "discourage casual download to desktop"
is very distinctly a stated goal of some contributors
to the discussion.  Others are not exactly in agreement
that such discouragement is logically possible.

A nice feature of the matrix I'm sketching here is 
that it highlights those controversies by exposing
very precise questions.  "If you say 'yes' to such
obstruction, does this column get a 'good' or a 'bad'?"
And there is a very finite number of such questions
to go through.   And so there should be a rule
that the "yes" proponents of a the question labeling
some row get to fill in the "yes-case" half of the
box, the "no" proponents the "no-case" half.

In the end, we can read off that matrix and have a 
list of "candidates for consensus".  What is a 
candidate for consensus?  Well:

For each row, choose a "yes or no" to the goal
stated for that row - then read down the corresponding
columns for each remaining technical recommendation
set.  If all entries in the column say "good"
then we have a candidate for consensus and a list
of goals achieved by that consensus.

I am agnostic as to whether or no any candidates
for consensus exist.  I thought I saw some but when
I asked pointed questions in support of them I was
accused of such things as trolling.

Nevertheless, such a matrix is exactly what I've
been working off of in formulating my questions and
comments - the only drawback/difference is that 
I've had to interpolate/guess, apparently offensively
incorrectly, how the yes or no proponents of various
rows would fill in the yes-case/no-case parts of each
box.

If we explicitly make and work off of the matrix - 
someone can volunteer to be secretary and make
a web page with a <table> - that can make the discussion
much more efficient and less heated.  We can quickly
identify what the real sticking points are rather
than getting caught up in personality conflicts.

-t

Received on Friday, 7 August 2009 07:13:47 UTC