🚀 go-pugleaf

rolfl@karl.uio.no (Rolf Lindgren) writes:


>Now, Web, I suppose, supports dynamic allocation of memory through
>Pascal's new() operator, whose behavior in extreme cases is unspecified
>and the execution speed of which is compiler dependent. 

>A similar problem is that if you use dynamic allocation of memory, you
>can't compare the speed of different implementations of a given algorithm,
>partly because the speed of different types of pointer allocation is
>compiler dependent and not easily computed. To be able to rank
>implementations of algorithms according to their speed is crucial if a
>program does a lot of operations on its data structure, as TeX does.

Sorry, /dev/null is full.
This is really nonsense.
There was difficulty with dynamic allocation in Pascal,
but there is absolutely no reason why it should not be introduced
into TeX in C.
In fact Karl Berry (in charge of UnixTeX distribution)
has said that he is working on this.

However, I doubt if it would be possible to go
from small to large TeX dynamically,
as the size of the elements in the large mem array 
(and other arrays) 
differs in the 2 cases.
For this reason,
it would still be necessary to have small and large TeX,
even if the latter was dynamically extensible.


-- 
Timothy Murphy  
e-mail: tim@maths.tcd.ie
tel: +353-1-2842366
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland