[lug] The deal with applets / notifications / panels / indicators / systray.....

[Rob Nagler]

On Sun, Feb 5, 2012 at 2:40 PM, Bruce Long wrote:
> Thanks for the thoughtful response.

Ditto. :)

> It sounds like you are recommending against using a pre-existing framework
> because it may not be tailored to your specific case.

Actually, I'm recommending against building a framework instead of
solving the problem.  If a framework emerges, so be it.  If, as you
solve the problem, a pre-existing framework fits the bill, go for it.
However, trying to solve a problem by creating a framework is usually
a mistake.

> Perhaps the pre-existing frameworks you have tried are simply not good
> enough?

I use frameworks designed to handle well-understood problems.  For
example, I use Linux, because we've known we can characterize the
problem of creating operating environments for programs.  Note that,
however, despite the fact we have been building distributed systems
since 1970, we haven't created a framework sufficiently well-suited to
solve the problem generally.    map-reduce, jQuery, and NFS are not
distributed systems.  They are trivial solutions to trivial problems
in distributed systems.

As a simple example of our inability to solve even the most
rudimentary problems in distributed systems (cache concurrency), if
you use gmail, go into one browser (iPhone), star a message, then go
into another browser (Chrome), star a message.  Why doesn't the star
show up in the first browser.  Even if you hit the refresh in the
Gmail app (not refresh on the browser itself), the star does not show
up.  Maybe it will for you, but it didn't for me just today.  I had to
refresh the entire application for it to contact "home base" for this
particular data point.

> Proteus was a pain to create. If there had been a better solution I would
> have gone with it.

I don't actually know how you know it is a good solution for your
problem.   I've been building a framework for applications development
for about 13 years now, and I'm still not sure if it is a good
solution for the problem, based on experience building ten or so
complex applications with it.  We've had, for example, three discrete
evolutionary steps to the presentation layer, and we are about to have
a fourth.  Browser technology has finally evolved to allow us to go
with a fully Javascript API.  When we began, Javascript didn't work,
and we had to limit page sizes to 10K or less.

Sorry for not addressing state vs. properties.  I think this issue was
clearly addressed by Turing and Church, and embodied in LISP, and
there isn't really much more to say about it. :)

Rob