[boost] Re: Re: Finite state machine requirements

From: Andreas Huber (spam2002_at_[hidden])
Date: 2002-09-05 19:44:26

Jeff,

> Not really.  One question though, have you looked at the FSM example
> that is part of the MPL paper:
> http://www.mywikinet.com/mpl/paper/mpl_paper.html

Yes, I had a look at the full source code which Aleksey was so kind to
provide. The approach has its limitations, as Aleksey pointed out:

> Interesting. I think that for small-to-medium FSMs a single STT is
> easier to maintain and understand (and probably check for
> completeness/correctness), as it keeps the FSM description "in one
> piece". However, your approach definitely makes sense to me as well,
> in particular because a single STT effectively limits the FSM size to
> something like 100-200 transitions (the compile times become
> inacceptable, or the compiler reaches its internal limits). If I were
> to design a full-fledged FSM framework, I would let both description
> methods co-exist.

http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&threadm=3d3abf94%40news.s
wissonline.ch&rnum=1&prev=/groups%3Fq%3DA%2Bbetter%2BC%252B%252B%2Bstate%26h
l%3Den%26lr%3D%26ie%3DUTF-8%26selm%3D3d3abf94%2540news.swissonline.ch%26rnum
%3D1

I tried to make the approaches coexist as Aleksey pointed out, but I have
yet to find a clean way to do this while supporting guards, hierarchical
states and concurrent states.

Regards,

Andreas