Re: [Gmod-ajax] Openlazlo

[Mitch S. <mit...@be...>]

On Fri, 2007-06-01 at 10:32 -0400, Michele Clamp wrote: 
> What's a div?

This may be overkill for answering a simple question, but it's a good
jumping off point for me to try and justify this approach.

A div is a plain-vanilla HTML element; it's just a rectangle that you
can position, style (change the background color/image, border
properties, etc.), and put stuff in.

Since features are often represented in genome browsers by rectangles, I
think divs are a pretty good fit for displaying them.  Like with
Simile::Timeline, there's a single axis (time for them, reference
sequence for us) along which we want to represent start/stop information
(events for them, features for us).  So most of the time a rectangle is
enough.

So in this context a div is a way of getting a web browser to draw
features on its own, rather than having the server draw features into an
image and send the image to the web browser.  Having the web browser
draw things gives us the flexibility to do lots of interesting things on
the client side.  Zooming with divs is sort of my pet example, but you
could also do things like directly edit features (i.e., click and drag
to adjust feature boundaries or create new features).

I think a lot of people consider SVG the Right Way of having a web
browser draw things, and it's true that divs are extremely limited in
comparison, but if you're dealing with things that live on a 1-d axis
then divs aren't so bad.  And I think people generally underestimate
what divs can do:
http://genome.biowiki.org/test/divbrowser/genome.html

My concern with SVG is that it's less consistently implemented in web
browsers.  IE doesn't have it at all, though it does have its own vector
drawing language (VML).  Some people are trying to make javascript
libraries that hide the differences between SVG and VML, but when I
looked they were relatively immature and limited in functionality, and
the ones I looked at presented an immediate-mode API ("draw a line from
here to here") on top of a retained-mode system ("there's a line from
here to here; remember it and draw it") which I think is backwards.

Maybe at some point SVG and/or VML will be the way to go; if we decide
to go in that direction we don't have to start from scratch.  The server
side can be exactly the same as with divs, and a lot of the client side
should also be the same.

I also think a lot of people see divs as being relatively large
heavyweight containers of things, rather than graphical elements in
their own right, so relatively few people have done things like
Simile::Timeline where the position/size of a div is used to convey
information.  And it's true, divs are fairly heavy, but they're
implemented in C++ and the browser writers put a fair amount of effort
into making them fast.  I think we've recently gotten to the point where
cheap, widespread hardware can use divs to do animation in real time.
And going forward I think the browser implementors are going to be
giving more consideration to this use case.

Mitch