[e-lang] Two-vat figure

Mark Miller markm at cs.jhu.edu
Tue Jun 21 11:19:49 EDT 2005


Ka-Ping Yee wrote:
> I've posted a new design for the two-vat figure at
> 
>     http://zesty.ca/2vat.png

Ping, this is good. I very much like the direction this is going.

The current draft of the paper will now be at 
<http://www.erights.org/talks/promises/paper/>. I put your draft 2vat diagram 
in as best I could, starting from the .png bitmap. Could you post the .svg 
sources? Thanks.

>     1.  The diagram is rearranged to read mostly left to right, instead
>         of bottom-right to bottom-left to top-right.  Reasoning:
>         left-to-right reading order is more natural.

Definitely a good move.


>     2.  The stack and queue are integrated into the structure of the vat.
>         Reasoning: they are a part of the way the vat works, so they
>         shouldn't look like other objects just floating around in the vat.

Good.

>     3.  The vats and objects are given non-cryptic names.  Reasoning:
>         words are easier to understand than abbreviations like XL.
>         (Mark, i know you said you preferred not to do this.  We can
>         change it back if you want, but i wanted to at least show what
>         i originally had in mind.)

Let's wait on the renaming issue - leave it this way for now. (In any case, 
even if we do keep readable names, these are the wrong names.)

There should be spaces around the "<-".


>     4.  The sequence of steps is labelled in a distinct style (white
>         on black filled circles).  Reasoning: the sequence of steps
>         should draw the eye.

Very good!


>     5.  Descriptions are added to steps 2, 4, and 5.  Reasoning: an
>         explanation helps show what is going on here; the message
>         names in the original figure aren't enough.

Good.


>     6.  Dotted arrows are added to show the eventual message being
>         placed on the queue and removed from the queue for processing.
>         Reasoning: without these arrows, it looks like the message is
>         being delivered immediately.  The arrows also make it clear
>         that the message arriving, the message on the queue, and the
>         message being delivered are one and the same message.  (In
>         the original there is no clear cue that the message on the
>         queue is the message on the incoming pointer.)

Very good.


>     7.  The number 33 is moved inside the message arrow.  Reasoning:
>         too many arrows produce visual clutter.  The 33 is a selfless
>         value, so the semantics are unchanged, the diagram is clearer,
>         and the messages containing 33 also appear more distinct.

I think that's good.


>     8.  The message arrows now point right instead of left.  Reasoning:
>         this makes more sense with the left-to-right reading order
>         since the messages travel from left to right.

Definitely the right thing.


>     9.  The shapes of the message arrows are simplified.  Reasoning:
>         this makes the diagram clearer by reducing the number of visual
>         elements and also makes more room for the contents (33).

I like that. I'll need to adjust the other figures to follow this convention. 
Is there any well-known name for this shape other than "arrow"?


> My goal here is to put enough information in the diagram so that even
> a casual reader whose eyes happen to land on the figure while skimming
> over the document has a decent chance of figuring out the story, and
> may then be enticed to read the details in the text.

Good.


> These 1-vat and 2-vat figures shade the stack and queue areas of the
> vat.  If it's really necessary for Springer the shading could be
> removed and replaced with partition lines, but shading is a better
> option when possible since the extra lines generate unnecessary visual
> clutter and line crossings.

I have no idea what's really necessary for Springer, but I just don't find the 
lines David put in (as seen in the current draft of figure 1) to add 
significant clutter. This minor degree of clutter removal isn't work risking a 
bad reproduction.

Historical: One reason I'm being strict about Springer's guidelines is that 
they basically destroyed the printed version of Paradigm Regained in an 
attempt to reformat it. The first clue I had that they were doing a wholesale 
reformatting of the document was when I received the paper proceedings. They 
obviously did a lot of guesswork along the way, but never once did they try to 
resolve a question by simply asking me.

-- 
Text by me above is hereby placed in the public domain

     Cheers,
     --MarkM




More information about the e-lang mailing list