Re: persistent ids/hashes from strings?

I would suggest that this XSLT application maintains a separate xml
document containing a unique element for every company.

Then, to give a hash to a company, it could use something as simple as
its position (count(preceding-sibling::company) in the companies' xml
document.

Whenever the lookup function fails to find a company (this means a new
company has been specified or a typo), the lookup function will create
a new Companies xml document, in which the missing company is appended
at the end and return it as part of its result.

The end result of the transformation will be the required one with the
addition of another file, containing the companies data, whenever new
companies have been added.

Cheers,
Dimitre Novatchev