Re: LL3 question

[Kimberley Burchett <address@hidden>]

On Thu, 8 Jan 2004, Neelakantan Krishnaswami wrote:

> A (late) question about LL3: I'm curious about what you think the most
> surprising talk was. By surprising, I don't only mean good, I mean
> which talk made you think about languages in a surprising new way.
> 
> When I was at LL2, the talk that surprised me the most was Oliver
> Steele's talk about LZX, the Laszlo language. The whole package was a
> surprise -- I was surprised to realize that other languages could
> target the Flash VM; I was surprised at how compact constraint-based
> descriptions of GUIs could be; and overall I was surprised at how
> "retrospectively obvious" their strategy was.

I'd have to say the Bluespec talk was the most surprising, in the sense
that you mean.  The way they separated the sophisticated static features
from the basic runtime features was unexpected -- it emphasized the
difference between these two aspects of programming to a much greater
degree than most languages I've seen before.  I'm tempted to say it felt
like they had designed a language wherein you pretty much had to use
macros to do anything interesting.  I expect that the constraints this
places on the language lead to all kinds of interesting and unique ideas.

If I hadn't already spent a lot of time thinking about dataflow languages
and how they can be implemented via compile-time expression rewriting, I'd
have said that the Father Time talk was the most surprising.  However, the
idea of thinking of expressions as functions over time, rather than as
static expressions that simply get re-evaluated on demand, was new to me.  
I had previously been thinking of dataflow more along the lines of the
spreadsheet model -- and if you think of it that way, things like the
delay function (which delays the propagation of a new value for a fixed
amount of time) simply don't occur to you.


Kimberley Burchett
http://www.kimbly.com/