On Thu, 29 Sep 2005 11:50 pm, Peter B. West wrote:
> Fopsters,
> I've always been somewhat sceptical of the new approach to page
> breaking, although I was prepared to concede that it would be a great
> achievement if you pulled it off.
> However, the closer the development has come to fruition, the more
> some of my original concerns have been reinforced.  Think about the
> enormous amount of intellectual effort that has gone into mapping the
> problem into Knuthian.  That effort is still under way.
> How is this going to be maintained?  Where are the Knuthian speakers
> who are going to do that job over the next few years?
> I'm surprised, in fact, that some of the old hands have not raised
> this question already.

I don't get it what you are aiming at here.

Are you saying that the Knuth approach to line or page breaking is 
inherently more difficult to understand and therefore harder to 

Apart from being one of the "best" (in terms of visual quality of the 
output) algorithms for breaking it is also IMO inherently simple. This 
is one of the beauties of many of Knuth's works. He is IMO a brilliant 
Computer Scientist who manages to solve complex problems using simple 
concepts and algorithms. The concepts and algorithms are also well 
documented in papers and books which are usually accessible through 
your nearest university library. Just take the "Breaking Paragraphs 
into Lines" paper. Yes, it is over 80 odd pages long but the important 
concepts are explained in the first 10 pages. In my case, when I delved 
"cold" into the fop layout code I had no idea what was going on, but 
after reading the initial part of the paper it all suddenly made sense.

So, where is the problem - Fop is using well documented concepts and 
algorithms to do its line and page breaking. Why should it be harder to 
maintain than some home cooked solution not backed up by previous 
research / papers / implementations (Tex)?

And if you take the recent discussions mainly driven by Jeremias 
regarding the bpd space resolution the core of the problem is not 
mapping it into the appropriate Knuth sequences. It is the 
implementation of the space resolution rules themselves, i.e. figuring 
out how much space to leave or not leave in a particular situation, 
which is the hard part. Generating the appropriate Knuth sequence once 
you know what the resolved space is is easy.

And quite a few of the stuff is also documented on the WIKI.

In summary, and I can speak here from my own recent experience, I don't 
share your concerns about the Knuth approach increasing the 
maintainability cost of the fop code base.

> Peter


Reply via email to