Hi Bill,

 
> A function interfaced to the xs module would be find, but how about an
> object approach?
> 

I feel more comfortable with the "classic" approach instead of the 
object approach. But these are just my feelings. Perhaps I am 
getting old...
This is my second perl module and I am making it in the same way 
I made my previous one. 
I think the more important component is the C library. This is the 
hard work. In fact, the xs module took just a few minutes.
Give me some days to get a decent library, then we can talk about 
the way to get the best perl implementation.

> 
> One thing that might be nice is that with this you could have separate
> modules for searching and for indexing.  Normally you might say "use
> SwishE::Search;", but saying use "SwishE::Index;"  or maybe just "use
> SwishE;" then you would get full function Swish in the same interface.
> 

At this moment, I am only working in the search part of swish-e.

> BTW -- would being able to use the index header as a place to store info be
> any use?  Could the index header be variable length?  I'm not sure what
> people would put there, but it might be cool to be able to make up
> additional headers when creating the index.  The index header format would
> have to be tagged or indexed, of course.

The index header is now of variable length. It consits of several 
lines staring with char "#" and the label of the variable. Personally I 
do not like this format because max length is assumed internally 
by the code. I would rather prefer somthing like:
<length-tag><tag-name><length-content><tag-content>
This is the storing way used for MetaNames, Propreties and Files.
This way is safer and faster and would make the code easier.

cu
Jose