cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Daniel Fagerström <dani...@nada.kth.se>
Subject Re: XMLForms and Flowmap
Date Thu, 13 Jun 2002 14:49:56 GMT
Diana Shannon wrote:

> 
> On Wednesday, June 12, 2002, at 06:31  PM, Daniel Fagerstrom wrote:
> 
>> Feel free to ask if you need further clarification about my integration
>> proposal.
> 
> 
> What *I'm* most interested in is a statement you made a while ago on 
> this list about the difficulties students can have learning about 
> continuations, from your direct experience teaching continuations. 


Why are continuations difficult?
--------------------------------

Continuations are a little bit like "goto on steroids". They are used in 
the study of denotional semantics which is a subarea of theoretical 
computer science. The aim of denotional semantics, is to describe (and 
prove properties of) programming constructs in terms of a very small 
functional language, (lambda calculus) that includes continuations. 
Continuations are extremely powerfull; program language constructs like 
assignement, goto, while, for-loops, throw-catch and coroutines to name 
a few can be defined in terms of a terse and cryptic, (but elegant, if 
you happen to like mathematical puzzles ;) functional program.

Some functional program langages, Scheme e.g. have continuations. A good 
use for continuations in such languages is to use them together with 
macros to define new language constructions. You can also use them for 
building a more intelectually demanding form of spagetti code ;)

Why can continuation based code be difficult to grasp? Even if most of 
us might not remember it anymore, learning to understand and use even 
the simplest programming language construct required a considerable 
effort once, allowing ad hoc introduction of new constructs forces us to 
repeat this effort often. Continuations forces us to think about the 
whole state of the program rather than about state changes. 
Continuations introduces non-local effects in the code, non-local 
effects are the main reason why gotos and exception handling can be hard 
to follow.

Flowmaps and continuations
--------------------------

In Flowmaps, continuations are used for implementing something quite 
similar to coroutines.

> Do you have any suggestions for the doc effort in this regard?


Focus on the coroutine aspect of the flowmaps, it is fairly easy to 
understand, while general use of continuations just isn't. For the 
flowmap user, the fact that flowmaps are implemented in terms of 
continuations should be of very limited importance otherwise we have a 
severe mix of concerns in the design.

I strongly sugests that we should _not_ make general use of 
continuations available in the flowmap language as there is no need for 
it for what flowmaps are intended to do: describing multipage flow in 
webapps.

> In what 
> aspects do students have problems: understanding the concept, applying 
> them to real-world situations, etc.? I've read up on continuations, 
> generally, but so many existing docs are tied to Scheme implementations 
> (which could frighten users, as has also been discussed). Can you point 
> to any existing teaching resources that you use? Care to donate anything 
> of your own, even a teaching outline? Any tips?


I tought denotatinal semantics, and the literature in that area is even 
more frightening than the Scheme literature;)

IMO we would make the effort to use flowmaps to high if we require 
people to learn continuations. I belive that it should be enough to 
teach about the coroutine like control structure, that happen to be 
implemented in terms of continuations. If it's not possible I think that 
is important feedback on that we have to work more on the basic concepts 
in the flowmap. I have a long unfinished RT where I compare FSM:s with 
the basics concepts in the Flowmap proposal and develop the coroutine 
analog a bit further. I'll check if it is in a comprehensible state and, 
if so, I'll send it to you.

I would also like to remind about some mails I wrote a couple of months 
ago where I explain why it's IMO is FS or even harmfull to allow local 
variables in the flowmap language: see 
http://marc.theaimsgroup.com/?l=xml-cocoon-dev&m=101861458122598&w=2 and 
some other mails in the "continuation fear" thread.

/Daniel


---------------------------------------------------------------------
To unsubscribe, e-mail: cocoon-dev-unsubscribe@xml.apache.org
For additional commands, email: cocoon-dev-help@xml.apache.org


Mime
View raw message